Populations were stratified into hierarchies of power landowners , knowledge priests and wealth traders. The most powerful — those who could raise the largest army — ruled. At the bottom of the hierarchy were women, servants and slaves. Women were regarded as the personal property of the father or husband and could be treated in whatever way the father or husband decided — basically women had no rights and could not hold property.

Health and healing in State Awareness focused on the use of traditional natural medicines and sacrifices to the gods — both animals and humans. Around 5, years ago, the leaders of city states started building empires: they invaded other states, plundered their riches and slaughtered their people.

Those they did not slaughter — mostly women — they captured and made into slaves to be sold in their home state to the highest bidder. From roughly 3, BCE to the 6th century AD, over 47 empires were created, most of which lasted for only two or three centuries. However, some, like the Roman Empire and the Chola Dynasty in India, continued for more than 1, years.

Over the next 1, years from the 6th century to modern times a further empires were created. Some of these empires, particularly those created from the 7th century to the 15th century, were driven from the worldview of Nation Awareness the spread of monotheistic religions , and others from the 16th century to the 19th century were driven from the worldview of Wealth Awareness establishment of colonies for commercial trade.

In the worldview of State Awareness, leaders have absolute power; they are feared by those around them and they consider themselves above the law. Fear is rampant in nations operating from the worldview of State Awareness, and caution is everywhere.

Leaders never feel safe. There is always someone in the wings waiting to grab power. For this reason, leaders in the worldview of State Awareness are extremely wary of those around them: there is intrigue and plotting everywhere.

Only the most powerful, the most fear-inducing and the most scheming leaders survive for significant periods. Leaders who operate from the worldview of State Awareness demand loyalty and require constant praise and adoration — they need their egos stroked.

They want what they want, when they want it, and you had better beware if you are unable to deliver it or if you cross them in any way.

Corruption is rife in the worldview of State Awareness. Leaders who operate from the worldview of State Awareness are usually men and, depending on their religion, either have several wives and many offspring or a succession of trophy wives.

In the worldview of State Awareness women are subservient to the needs of men. They are little better than servants or slaves and are often treated as sexual objects.

Leaders in the worldview of State Awareness take pride in displays of military strength and accomplishment. They need such demonstrations to feed their self-esteem: they want to feel powerful and they need to show off their strength. They may even resort to displays of perceived manhood or by making public appearances with beautiful women.

They like to father numerous male children. Female children are less welcome. Leaders operating from the worldview of State Awareness will lie, cheat and manipulate to become top dog.

Corruption and bribery are everywhere. This is the worldview of despots and dictators and all those who use fear to manipulate others to gain power. F frontal region, T temporal region, P parietal region. We additionally compared the classification performance of ECI awa among patients with DoC using TMS—EEG data when we excluded electrodes in different brain regions from the input during classification.

Consequently, AUC was 1. We show that ECI clearly distinguishes between low and high states of arousal and awareness in TMS—EEG results under sleep, anesthesia, and patients with DoC. Our results suggest that this proposed indicator could similarly be used for resting-state EEG data without TMS under anesthesia and patients with DoC, yielding the same degree of accuracy.

In addition, a high correlation with PCI, which measures the integrated EEG response of the thalamocortical system to a direct perturbation induced by TMS 16 , proves that ECI awa is reliable using TMS—EEG data.

It also shows that the two measures calculated entirely independently using different methods resulted in the same conclusion, which is a sign that deep learning is indeed a valid and reliable approach. For an ML-based indicator, ECI can be calculated using very few trials.

Furthermore, because the classifier learned specific features of the data on its own, our indicator can be computed regardless of whether TMS is applied or its location. Therefore, ECI is a significantly practical and reliable indicator to evaluate levels of consciousness under various conditions. Our analyses using LRP highlighted the major role of the parietal region in determining consciousness, as the classifier primarily uses brain activity in this lobe for predicting low and high states of arousal and awareness.

TMS—EEG responses under sleep exhibited well-known phenomena In wakefulness, TMS generates a series of low-amplitude high-frequency activities related to cortical flow in long-range connections A similar long-lasting response is evoked during REM sleep with subjective experience During NREM sleep with no subjective experience, TMS triggers larger, low-frequency activity that quickly dissipates 32 , which is the hallmark of bistability in the thalamocortical system.

Cortical effective connectivity is also broken down during NREM sleep 33 , The brain response to TMS perturbation was already used to distinguish the levels of awareness, irrespective of sensory processing and motor responses under physiological, pharmacological, and pathological conditions 34 , 35 , Therefore, this TMS-evoked response was applied to our end-to-end CNN framework.

Similar to several studies using ML, we observed higher two-class low or high classification accuracy using CNN when compared to linear discriminant analysis LDA and support vector machine SVM for both arousal and awareness. This suggests that our framework is especially relevant for EEG results, which possess several nonlinear features.

The classification performance of spatiotemporal information was higher than that of spatio-spectral information. Using our framework, it was shown that temporal information discriminates different levels of consciousness more clearly than spectral information, as the functional connectivity associated with consciousness changes in both space and time However, this does not imply that temporal information is more important than spectral information for distinguishing consciousness.

Temporal information has more distinct characteristics than spectral information for predicting the state of consciousness in the proposed framework. Nevertheless, PCI also used spatiotemporal dynamics in TMS-evoked responses 16 , which is significantly important for distinguishing consciousness.

We applied transfer learning to a single domain as well as multiple domains. In the sleep domain, classification performance was high when trained on only the sleep domain or together with the anesthesia domain.

Due to a large number of trials in the sleep domain, training was performed satisfactorily using only sleep data. The distance of the averaged TMS-evoked potentials under the domains of anesthesia and sleep was significantly close, which indicated that the two domains were highly similar Thus, a close distance implies that the two domains have similar patterns, and the classification performance indeed increased when these domains were trained together.

In the anesthesia and DoC domains, the trials of a single domain were not sufficient; thus a higher classification performance was achieved when trained with similar domains. In addition, brain signals change over time, even when recorded with the same participants because of physiological and psychological differences over time Therefore, participant-independent learning such as the LOPO cross-validation is considerably difficult, as opposed to participant-dependent learning We solved these problems via transfer learning utilizing multiple domains and participant-independent ECI.

ECI, by averaging interclass probability, distinguished whether each state was low or high in both arousal and awareness. Because of its inherently high two-class performance, it was possible to calculate the discriminable ECI using a few trials in a single session.

The criteria used for determining ECI aro and ECI awa were different, especially for REM sleep, ketamine-induced anesthesia, and UWS patients. Although the dataset was the same, these states were trained by different labels depending on arousal and awareness.

For instance, REM sleep is high in awareness but low in arousal. The proposed classifier learns models for these different criteria by training itself based on the criteria of arousal and awareness. Specifically, because the learning ability of the CNN is derived from the automatic extraction of complicated representations from EEG signals 39 , it can properly distinguish between both states, even if the same state has different labels depending on the criteria.

According to the LRP, we observed higher relevance scores in the parietal regions, compared to other regions at the scalp level. This was observed for all data, including TMS—EEG and resting-state EEG results.

The brain regions that led to decision-making were similar in arousal and awareness. However, this does not imply that arousal and awareness are supported by the same underlying neurophysiological mechanisms.

The relevance scores simply explain the patterns of cortical EEG activity resulting in this classification. In sleep and healthy wakefulness, when arousal and awareness were high, they were highly relevant in the parietal region.

This EEG feature, which distinguishes high and low states in arousal and awareness, can be interpreted in line with the posterior hot zone of consciousness Local changes in this parietal region are associated with the occurrence of dreaming and unconscious sleep 41 , 42 , and our framework may learn from EEG pivotal features recorded in this area.

The importance of the posterior hot zone has already been emphasized using the within-state sleep paradigm 43 , Similar to sleep, we observed high relevance scores for the parietal region in the domains of both anesthesia and patients with DoC.

This implies that EEG activity in this region had a decisive effect in determining the high and low states of arousal and awareness.

The increased slow-wave activity was observed under propofol- and xenon-induced anesthesia when compared to healthy wakefulness before drug administration In addition, just as cortical neurons induced bistable changes during NREM sleep, TMS in propofol-mediated anesthesia-induced low-amplitude, low-frequency, positive—negative potentials, and TMS in xenon-mediated anesthesia caused a significantly large amplitude but stereotyped positive—negative deflection Moreover, under ketamine-induced anesthesia, TMS-evoked response was determined to be similar to REM sleep, which features dreaming during a low state of arousal Previous studies have shown that the change in slow waves induced by propofol is primarily observed in the posterior hot zone 45 and the posterior main hub is disrupted during anesthesia-induced alteration of consciousness 7.

In UWS patients, TMS triggered a local and slow response similar to NREM sleep and general anesthesia, whereas MCS patients showed complex TMS-evoked responses Similarly, differences in alpha connectivity between the UWS and MCS patients are apparent within the posterior hot zone That is, when determining whether arousal and awareness are low or high, our classifiers used differences and changes in EEG activity over the parietal region to make decisions.

Particularly, similar results were observed in patients with DoC using several TMS target sites, and in sleep and anesthesia domains, where the parietal region was primarily stimulated.

This suggests that our findings regarding the parietal region are unrelated to the TMS target site. Thus, our trained model used neurophysiological features to classify whether arousal and awareness are low or high.

This indicates an appropriate design of our model as the classification decision was primarily based on EEG signals over the parietal region, which is suggested to be a hot spot of consciousness, compared to the frontal region The difference in the parietal region could be clearly identified through correct trials during sleep and healthy wakefulness.

It is meaningful that the frontal region contributed less than parietal regions in the context of the controversy regarding the spatial localization of the neural correlates of consciousness 40 , Considering the subcortical influences related to striatal-thalamic circuits, it has been recently observed that the parietal region contributes more to the levels of consciousness than the frontal region The implication of the parietal cortex in consciousness has also been demonstrated in other neuroimaging modalities, such as functional magnetic resonance imaging 48 and magnetoencephalography This study does have certain limitations.

First, our sample size was relatively small. In the future, it will be necessary to further test the reliability of the proposed indicator with larger cohorts and validate it at the clinical level before implementing it in a clinical setting.

Sleep experiments would also have to be applied to more participants in the future. Second, we explored the possibility of calculating ECI with a minimal amount of data, up to a single trial.

However, we did not attempt to measure ECI in real time. Thus, in the future, ECI could be calculated in real-time for practical application. Third, ECI does not differentiate between physiological, pharmacological, and pathological conditions, but distinguishes between high and low states of arousal and awareness.

ECI can thus distinguish between REM sleep or ketamine and wakefulness. It may also be difficult to select the model to use when calculating ECI since the domain has to be known beforehand.

Nevertheless, if a single domain has sufficient trials, the LOPO approach would be the most accurate. Another limitation might be related to the possible contamination of TMS-EEG data by auditory and somatosensory components.

As in previous studies 17 , 35 , 36 , to avoid auditory and somatosensory co-stimulation, participants wore earphones with noise masking and a thin foam between the scalp and the TMS coil that was used.

Although it is difficult to systematically rule out the contribution of sensory co-stimulation in every measurement, the application of effective noise-masking procedures 50 and the real-time monitoring of data quality during the acquisition 51 may significantly mitigate this issue Finally, ECI indicates whether arousal and awareness are low or high and cannot be considered functional.

It, therefore, should be developed into a functional index. In conclusion, we proposed ECI as a neurophysiological indicator to simultaneously discriminate the levels of arousal and awareness in modified states of consciousness. This tool allows disentangling the levels of consciousness, with a single measure, in different clinical settings such as monitoring surgical interventions i.

This indicator was validated under different physiological, pharmacological, and pathological conditions, and it reliably disentangled low levels from high levels of both arousal and awareness.

Besides, the proposed ECI is considerably accessible and practical, as it can be applied to resting-state EEG without TMS, and requires fewer trials. Therefore, the proposed indicator can be a reliable discriminator and valuable tool as an objective measure of consciousness.

As parietal regions appear to be the most relevant for classification, an EEG configuration around that area could be sufficient if ECI is used in clinical practice.

These findings could be useful in diagnosing severely brain-injured patients and monitoring their levels of consciousness in real-time, especially in clinical settings where time constraints preclude long-duration assessment.

The proposed reliable ECI can provide insights into the classification of conscious levels using deep learning and neural correlates of consciousness. The sleep dataset included six healthy participants five males, aged The inclusion criteria included i between 18 and 75 years of age and ii in good general health.

The exclusion criteria were as follows: i neurological, psychiatric, mood, and sleep disorders, ii contraindications for TMS e. All participants provided written consent, and the experimental paradigm was approved by the Institutional Review Board IRB at the University of Wisconsin—Madison HSC Supplementary Table 8 lists the TMS target site and the number of sessions and trials.

The participants were awoken by an alarm sound that lasted for 1. They were then asked if they had had conscious experience. The TMS—EEG experiments were performed over a period of four or five nights per participant.

The anesthesia data were previously published by Sarasso et al. The inclusion criteria included i older than 18 years and ii stability of vital parameters. The exclusion criteria were as follows: i neurological, cardiovascular, psychiatric, and mood disorders, ii contraindications for TMS e.

TMS was applied over the left parietal or motor regions after participants reached deep unresponsiveness a score equal to 5 in the Ramsay scale, which corresponds to no response to external stimuli following standard anesthetic procedures The stimuli target site and the number of trials are listed in Supplementary Table 9.

In addition, upon waking up from anesthesia, reports about the conscious experience during anesthesia were collected.

Conclusively, the participants reported little conscious experience during propofol- and xenon-induced anesthesia, but vivid dreams were experienced during ketamine-induced anesthesia.

For patients with severe brain injury, the data of six UWS patients 2 males, 4 traumatic brain injuries, time since the injury of All of them fell into a coma due to brain injury and presented a prolonged state of impaired consciousness.

The inclusion criteria included i older than 18 years and ii diagnosis of DoC following a severe acquired brain injury. The exclusion criteria for patients were as follows: i patients having significant neurological, neurosurgical, or psychiatric disorders prior to the brain injury that leads to DoC, ii patients having any contraindication to TMS—EEG or magnetic resonance imaging electronic implanted devices, active epilepsy, external ventricular drain , and iii patients who were not medically stable.

Accredited experts performed repeated CRS-R for each patient, including on the day of the TMS—EEG examination and before the fluorodeoxyglucose-positron emission tomography FDG-PET scan.

TMS—EEG data were acquired similar to previous studies 16 , Moreover, added data were newly included as follows: 9 UWS patients 6 males, 5 traumatic brain injuries, time since the injury of 6. These added data were recently acquired by Dr. Steven Laureys team at the University of Liège.

For the added data, the inclusion and exclusion criteria for patients were exactly the same as the previously used data. These data are part of a bigger study conducted in the frame of the Human Brain Project.

The detailed demographic and clinical information of severely brain-injured patients is listed in Supplementary Table The TMS target site was selected using a neuronavigation system over the parietal, motor, or premotor regions, avoiding structural lesions using the magnetic resonance imaging data of the patient.

The stimuli target sites for all participants are listed in Supplementary Table The participants remained awake or were kept awake using the CRS-R arousal protocol in between TMS stimulation During all the sessions, earphones presenting white noise were used to reduce the noise of the TMS pulses and we used a thin foam between the scalp and the TMS coil to avoid somatosensory evoked potentials.

These pulses were presented at random intervals of 2—2. In particular, the TMS stimulation was always performed in the medial half of one hemisphere, to avoid any muscle artifacts. Finally, we used resting-state EEG without TMS for the domains of anesthesia and severely brain-injured patients.

These data were acquired using the same participants as in the TMS—EEG experiments. Finally, among previously published data, the data with a signal-to-noise of 1. TMS—EEG data were preprocessed using the SiSyPhus Project MATLAB program University of Milan, Italy and the EEGLAB toolbox The signals were down-sampled to Bad channels were manually detected and interpolated using superfast spherical interpolation for artifact removal.

The data were re-referenced to an average reference 58 , Resting-state EEG data were processed using the EEGLAB toolbox The number of trials for each session we used is listed for patients under anesthesia Supplementary Table 12 and severely brain-injured patients Supplementary Table More details related to this deliberate choice are reported in Supplementary Note 2.

In the first step, EEG data were converted from 2D raw signals to 3D input. To preserve the spatial information and characteristics of EEG, we used spatio-spectral and spatiotemporal 3D features.

n indicates the number of electrode channels. However, these simple signals do not capture all the spatial information characteristics in the brain. Therefore, we converted 1D data vectors to 2D EEG data meshes using the spatial information of the electrode location.

Zero was inserted in the place of a null electrode in 2D matrices at time index t Finally, we calculated 3D data by adding spectral or temporal information Supplementary Fig. Specifically, spectral information was divided into 5 frequency bands: delta 1. Step 2—Calculation of interclass probability using CNN: The model was trained to distinguish both arousal and awareness in terms of whether they were low or high.

We first calculated the domain similarity based on the cosine distance for domain transfer learning. The cosine distance D between domains A and B is defined as follows:. This similarity was used when selecting the source domain for domain transfer learning In calculating the similarity according to the states of arousal and awareness, the labels in each state are different.

During sleep, NREM sleep with no subjective experience and REM sleep with subjective experience were learned as low arousal, whereas healthy wakefulness was learned as high arousal. Conversely, in awareness, REM sleep with subjective experience and healthy wakefulness indicated by open eyes were learned as high awareness, while NREM sleep with no subjective experience was learned as low awareness.

Specifically, we used stage 3 of NREM sleep for clear feature extraction of deep sleep. Under general anesthesia and in healthy wakefulness before anesthesia, the arousal state was divided as follows: i low, when under ketamine, propofol, and xenon-induced anesthesia, and ii high, when in healthy wakefulness before anesthesia.

Conversely, the awareness state was divided into low under propofol and xenon-induced anesthesia and high under ketamine-induced anesthesia and healthy wakefulness before anesthesia.

Finally, in patients with DoC, UWS and MCS patients were distinguished in terms of awareness: low UWS patients and high MCS patients. The UWS and MCS patients corresponded to high arousal. Deep learning was conducted in a MATLAB environment powered by a TITAN V GPU.

We used the LRP toolbox 30 for CNN classification and interpretation. The CNN was applied to the two components of consciousness arousal and awareness.

In each architecture, we inserted five convolutional layers with 2D filters for the deep neural network. Finally, the generated feature maps were flattened into a 1D vector. A softmax layer was used for classification.

In the softmax layer, each element indicates the probability that the original input belongs to the corresponding class. In this training procedure, the parameters in the deep neural network were learned through back-propagation.

The activation function in each convolutional layer was a rectified linear unit. The detailed CNN architecture is presented in Supplementary Table The Adam optimizer was used with an initial learning rate of 0.

Specifically, we used hyperparameters as values of 0. The batch size was 25 for training, and the maximum number of training iterations was five times the number of training data. Consequently, the output of this architecture was the probability of each class.

For a comparison with other classifiers, we also considered an LDA classifier 62 and SVM with polynomial kernel function 63 using the same input data as fair baseline methods. The classification performance was measured in LOPO-nested cross-validation for the generalized neural network. This method is a special case of k -fold cross-validation.

Specifically, all participants except one the target participant were used for training, and the target participant was then tested using the classifier. This process was repeated for each participant. Further, internal validation sets inner cross-validation were performed to choose the hyperparameters of the model Thus, the same hyperparameters were selected in the external validation sets The LOPO cross-validation procedure uses data efficiently and can reduce overfitting.

It also provides unbiased estimates of the averaged classification error for all possible training sets The same LOPO-nested cross-validation is applied in LDA and SVM using the Berlin brain—computer interface toolbox Three possible values 0.

For spatio-spectral input and spatiotemporal input 1 and 0. Step 3—Calculation of ECI in a single session: We obtained each interclass probability according to two components arousal and awareness at the previous step.

In each TMS session, the interclass probability was averaged to calculate an ECI. This approach has the advantage of being able to offset outliers. N is the number of trials in a single session j. Consequently, ECI is expressed as a 2D indicator representing both arousal and awareness simultaneously.

As mentioned earlier, we used the LOPO cross-validation. It is to be noted that only data from one participant were used as the test, and the data from the remaining participants were used for training. Finally, the test and training data did not demonstrate an overlap at all.

Step 4—Interpretation using LRP: We used LRP based on a backward-propagation mechanism for interpretability of the deep neural networks. This calculated the pixel-based decomposition process This score is calculated through the backward propagation of the model input.

Therefore, relevance scores describe a single nonlinear decision for the output corresponding to each input Through this method, we can observe not only the interpretation of classification decisions but also what features the model has learned To investigate which brain regions and signals caused these classification results, we compared the relevance scores from the LRP by dividing them into the following three regions 69 : the frontal Fp1—2, Fpz, AF1—2, AFz, F1—8, and Fz , temporal FT9—10, T7—8, TP9—10 , and parietal CP1—6, CPz, P1—4, P7—8, and Pz regions.

We focused on the frontal, temporal, and parietal regions when comparing the relevance score resulting from the LRP as there is an ongoing debate regarding which brain area, i. We also included the temporal region as activation in the NREM sleep increases in this region Additional EEG analyses are shown in Supplementary Methods as follows: i performance according to the number of trials in the ECI calculation, ii comparison of the difference between correct and incorrect trials, and iii classification performance using EEG signals excluding frontal or parietal electrodes.

We compared ECI awa with the PCI values computed following the same procedure described in PCI measures the complexity of the spatiotemporal patterns of cortical activity significantly evoked by TMS PCI ranges between 0 minimum complexity and 1 maximum complexity.

The Kruskal—Wallis test was also performed to compare the classification performance using transfer learning. To investigate the discrimination of ECI in each state of consciousness, the feedforward network was trained with 20 hidden layers using the LOPO approach.

For each output class, the AUC, sensitivity, and specificity were calculated using ROC analysis. The Kruskal—Wallis test was employed to investigate if there were any differences in relevance scores among brain regions from LRP under sleep.

In addition, the Kruskal—Wallis test was performed to explore the differences in relevance scores from LRP under the condition of anesthesia and for severely brain-injured patients.

Further information on research design is available in the Nature Research Reporting Summary linked to this article. All data sleep dataset, anesthesia dataset, already published brain injury dataset, and new added brain injury dataset generated and used in this study have been deposited in a local database and are available upon reasonable request to Olivia Gosseries.

The sleep dataset was previously reported by Nieminen et al. The anesthesia data were also previously published by Sarasso et al. The published brain injury dataset was previously reported by Bodart et al.

However, new added brain injury dataset has not yet been published. The raw EEG data are protected and are not made publically available owing to data privacy laws, but are available from the corresponding author upon reasonable request.

Source data are provided with this paper. Sanders, R. Unresponsiveness unconsciousness. Anesthesiology , — Article PubMed Google Scholar. Darracq, M. et al. Evoked alpha power is reduced in disconnected consciousness during sleep and anesthesia.

Article ADS PubMed PubMed Central Google Scholar. Colombo, M. The spectral exponent of the resting EEG indexes the presence of consciousness during unresponsiveness induced by propofol, xenon, and ketamine.

Neuroimage , — Article CAS PubMed Google Scholar. Lendner, J. An electrophysiological marker of arousal level in humans. eLife 9 , e Article CAS PubMed PubMed Central Google Scholar. Mashour, G. Neural correlates of unconsciousness in large-scale brain networks.

Trends Neurosci. Casarotto, S. Exploring the neurophysiological correlates of loss and recovery of consciousness: perturbational complexity in Brain Function and Responsiveness in Disorders of Consciousness ed Monti, M. Bonhomme, V. General anesthesia: a probe to explore consciousness. All it takes is to practice the methods of meditation noted above.

If you practice the right edition where you reach the state of pure consciousness, you will soon be able to perceive the shifting of these states and discover that this gap is the 4th state of pure consciousness. This 4th state undergirds the other default states and is a critical element in the shift to the higher state of awareness known as witnessing.

Here is where we get another answer to the question: what is expanded awareness? The 4th partition of awareness shows us that we have other levels of perception available which do not depend upon the five external senses. The theory of expanding consciousness proposes the idea that this expansion has no limits.

The states of awareness we describe in this article are only the beginning. Reaching higher states of consciousness is a goal at the heart of many ancient cultures. The first spiritual process is the journey into consciousness, called the Shamanic Journey. It produces a unique state using rhythmic sound and imagination.

Michael Harner 4 , a noted anthropologist, writer, and Shaman , calls this the Shamanic State of Consciousness SSC. It is the fifth primary partition of consciousness. The brainwaves are in the theta-wave bandwidth while in this partition.

We find this awareness-altering process in many cultures around the world. This process has a common theme across many indigenous cultures. While in this state, you use creative visualization techniques to project your awareness in an imaginary landscape.

You are both the director and participant within a non-ordinary reality landscape, which parallels lucid dreaming. The Aboriginal tribes of Australia have a peculiar variation of the Shamanic journey they call Dreamtime. This technique allows them to observe events in other places and times.

In essence, they found the secret to time travel. This partition of consciousness is significant for several reasons. First, it combines two partitions of consciousness, the 4th state with the waking state, to create a new partition of consciousness.

Second, it proves we can expand awareness to view reality from two different viewpoints. So, we learn our consciousness unbound.

It can move beyond the confines of our bodies. It is another stepping stone to other higher perception levels. It expands awareness to perceive reality from two different vantage points simultaneously. One achieves this 6th state of consciousness called witnessing through regular Japa or TM meditation.

This process draws the mind naturally to the state of bliss consciousness. So, it is natural to bring this quality into the waking state of ordinary reality. The 6th state is where we observe our experiences from a unique vantage point without getting attached to them. At this level of awareness, we can observe external experiences, thoughts, and emotions without identifying with them.

When we enter this state, we detach from the limiting beliefs and biases that usually hold us back. Those who reach this partition describe it as a state of witnessing. One can perceive reality from two vantage points at the same time.

Some describe seeing themselves from above, others from behind or alongside. They also perceive reality with their body but are witnessing themselves simultaneously from another point. Some refer to this experience as astral projection. This partition of consciousness provides experiential proof that what we perceive as ordinary reality is more aptly a collective holographic construct.

You can feel the distant rumble of thunder on the horizon, yet you have no idea of the deluge your life is about to experience. What are you? You are the Spirit. This sixth partition of consciousness is a major step in the evolutionary progression toward what we call enlightenment.

Believe me, you will not regret it. It is possible to get glimpses of the 6th state of consciousness, the state of awareness called witnessing, when you awaken from a deep sleep.

In those precious moments of stillness between sleeping and waking states, it is possible to become aware that you are sleeping.

Furthermore, the enhancement of dynamic functional connectivity, through the phase modulation at low frequency, between the PFC and other brain regions in the early stage of the awareness trials may explain the mechanism of conscious access.

These results indicate that the PFC is critically involved in the emergence of awareness. This paper reports valuable results regarding the potential role and time course of the prefrontal cortex in conscious perception.

Although the sample size is small, the results are convincing , and strengths include the use of several complementary analysis methods. The behavioral test includes subject report such that the study does not allow for distinguishing between phenomenal awareness and conscious access; nevertheless, results do advance our understanding of the contribution of prefrontal cortex to conscious perception.

A conscious state means that a person has some kind of subjective experiences, such as seeing a painting, hearing a sound, thinking about a problem, or feeling an emotion 1.

The neural mechanism of consciousness is one of the most fundamental, exciting and challenging pursuits in the 21st-century brain science 2. Among the various aspects of consciousness, visual awareness, as a core of consciousness, has attracted enormous attention in consciousness research.

Visual awareness seems quite simple: we see an object and seem to immediately know its shape, color and other properties.

However, the neural processes behind the emergence of visual awareness are quite sophisticated and largely unknown 3. The typical method to study the neural mechanism of visual awareness is to present external stimuli with similar physical properties but to elicit different conscious experiences, such as using near perceptual threshold stimuli, visual masking, or attentional blinks 4 , 5.

Then, the aware and unaware states are determined through the subjective reports of participants, and the awareness-related neural activity is dissociated by the brain activation difference between the two states.

Using this approach, some studies have found that activities in broad brain regions are correlated with visual awareness. For example, an fMRI study of lexical visual masking found that only consciously perceived and reported words can induce activation of higher-order cortical regions such as the frontal and parietal lobes 6.

Another study using a visual masking task found changes in the amplitude of intracranial electroencephalography iEEG signals and high-gamma power in occipital-temporal regions, which indicated that these regions played an important role in the emergence of awareness 7.

However, there is still much debate about the origin of conscious awareness in the brain. Some hypotheses global neuronal workspace theory, higher-order theory, etc.

argued that visual awareness arose from the frontoparietal network, in which the prefrontal cortex played a key role, and they argued that the earliest biomarker of awareness in EEG is the P3b wave onset approximately ms after stimulus onset 8 , 9. For example, based on the electrophysiological study of a visual awareness-guided saccade selection paradigm in macaques, by comparing the action potential firing characteristics of neurons in the dorsolateral prefrontal cortex dlPFC , V4 and V1, only dlPFC neurons were closely associated with visual awareness states.

Therefore, the authors argued that stimuli must elicit a threshold level of prefrontal activity for conscious detection to occur In contrast, other hypotheses integrated information theory, recurrent processing theory, etc.

This controversy is partially caused by the confounding effect of report-related activity, such as motor planning and execution, on awareness-related activity.

The reason is that the study of awareness usually requires subjective reports to determine different awareness states, while the prefrontal cortex is closely related to such report-related functions To solve this problem, some researchers use the no-report paradigm to rule out report-related impacts.

For instance, a functional magnetic resonance imaging fMRI study employing human binocular rivalry paradigms found that when subjects need to manually report the changing of their awareness between conflict visual stimuli, the frontal, parietal, and occipital lobes all exhibited awareness-related activity.

However, when report was not required, awareness-related activation was largely diminished in the frontal lobe but remained in the occipital and parietal lobes Nevertheless, the no-report paradigm may overestimate the neural correlates of awareness by including unconscious processing, because it infers the awareness state through other relevant physiological indicators, such as optokinetic nystagmus and pupil size In the absence of subjective reports, it remains controversial regarding whether the presented stimuli are truly seen or not.

Therefore, neither the typical report paradigm nor the no-report paradigm is ideal for visual awareness research, thus, it is difficult to solve the problem of whether the PFC participates in conscious perception. In addition, when studying the role of the PFC in visual awareness, the noninvasive technologies mostly used in previous studies, i.

Although fMRI research can detect the whole brain activation pattern, it is difficult to describe the dynamic changes and sequences of different brain regions in the rapid process of visual awareness due to its limited time resolution approximately several seconds.

For example, even though a recent fMRI study using the no-reported paradigm found that there are still visual awareness-related activities in the PFC 17 , it remains unclear when these activities begin and whether PFC activity occurs earlier or later than other brain regions.

At the same time, although EEG and MEG have a very high temporal resolution, their low spatial resolution and low signal-to-noise ratio limit their interpretation of awareness-related activities, especially high-frequency activities highly correlated with local population neuronal activities.

For example, although increasing evidence shows that the earliest biomarker of visual awareness in EEG is more likely to be the VAN, which starts at approximately ms may be delayed to ms for low-visible stimulus and peaks at the parieto-occipital region 18 — 20 , rather than the P3b wave onset at approximately ms which peaks in the central frontoparietal region, it is still unclear whether there are local visual awareness-related activities in the PFC during the VAN period.

In addition, a few recent studies have suggested that awareness may be related to the activities of brain networks, including the PFC, rather than that of a single brain region.

However, most of these brain network studies used noninvasive research methods 17 , 21 , and due to the limitations mentioned above, it is difficult to detect the dynamic changes in brain networks in a more precise spatial-temporal profile during the generation of visual awareness.

To address the questions mentioned above, we designed a novel visual awareness task Figure 1 that can minimize motor-related confounding, while retaining the explicit subjective report.

The contrast of the cue stimulus grating was roughly divided into three levels: well above the visual perceptual threshold, near the visual perceptual threshold and zero contrast no stimulus. Subjects were required to choose a saccade direction according to the visual awareness state seeing the grating or not seeing the grating and the color of the fixation point.

Notably, subjects were unable to choose the saccade direction or effectively prepare for saccades until the color of the fixation point was changed. Moreover, the subjects needed to choose the saccade direction, both in the awareness and unawareness states, according to the color of the fixation point so that the report behaviors were matched between the two awareness states.

Such a paradigm can effectively dissociate awareness-related activity from motor-related activity in terms of time the fixation point changes color ms after the grating onset and report behavior, thereby being report-independent of the awareness states 22 , thus minimizing the impact of motor-related factors on visual awareness.

In addition, saccade parameters such as reaction time, etc. under different awareness states can be compared to explore the influence of visual awareness on saccade behavior. Then, we applied the above behavioral paradigm to 6 clinical patients with implanted electrodes stereoelectroencephalography, sEEG in prefrontal and other cortices and 10 healthy subjects.

Schematic diagram of the visual awareness task. A trial started when a fixation point 0. After the subject fixated on the fixation point for ms, a cue stimulus Gabor grating, 2×2° circle was presented for 50 ms at a fixed position 7° on the left or right, see Methods side of the screen for all participants.

After another ms delay, the color of the fixation point turned red or green, and two saccade targets 1×1°, white square appeared at fixed positions 10° on the left and right sides of the screen. If the grating was seen, the green fixation point was required to make a saccade to the right target, while the red fixation point was required to make a saccade to the left target.

If the grating was not seen, the rule of saccadic direction was inverted. To detect the effect of cue stimulus grating contrast on the emergence of visual awareness, we first analyzed the proportion of participants being aware of the grating as a function of its contrast.

Figure 2A shows exemplified sessions from a patient and healthy subject, and Figure 2B shows the population results of the two participant groups. Such results showed that the grating contrast level had a direct impact on the emergence of visual awareness and thus is a reliable way to study visual awareness.

In addition, in trials with no grating and grating contrast well above the perceptual threshold, subjects showed high accuracy patients Psychometric detection curve in a single session.

Left panels from one patient and right from one healthy subject. Each black point in the graph represents the aware percent in a contrast level, and the black curve represents the fitted psychometric function.

Psychometric detection curves for all participants. N, number of subjects; R 2 , coefficient of determination. Bars show the mean values of saccadic reaction time. In the near-threshold condition, the mean reaction times of patients and healthy subjects in the aware trials were The mean reaction times in the unaware trials were Each black dot represents one participant.

The gray solid line represents the paired dots. Error bars represent the standard error of the mean SEM. Left, right and top views of all recording sites projected on an MNI brain template.

Each color represents a participant. In all brain images, right and up side of the image represent the right and up side of the brain. We compared the saccadic parameters between the awareness and unawareness states in near-threshold contrast levels NA vs.

NU , as well as in high and low contrast levels HA vs. LU , as a control. While saccadic reaction time showed a significant difference between the two awareness states Figure 2B , the other parameters did not.

Figure 2C shows the mean saccadic reaction time of patients and healthy subjects under different awareness states and contrast levels. The results showed that the saccadic reaction time in the aware trials was systematically shorter than that in the unaware trials. Such results demonstrate that visual awareness significantly affects the speed of information processing in the brain.

While the patients were performing our visual awareness task, we recorded the local field potential LFP from recording sites Figure 2D , projected on Montreal Neurological Institute MNI brain template ICBM , in which sites were located in the PFC.

The table in Figure 2D shows a summary of the number of recording sites and patients in different lobes and structures. While looking through the event-related response of the LFP data, a striking phenomenon in all patients was that the amplitude of LFP showed remarkably different patterns between awareness and unawareness states.

We show the grand average left and single-trial data right of LFP activity of recording sites in the PFC from each patient in Figure 3A , which represents the typical awareness-related neural activity of the PFC.

There was a vigorous visual evoked response under the HA condition light red line compared to the LU light blue line condition. LFP activity of example sites from each patient.

Each panel represents one example site from one patient. In each panel, the left plot shows the grand average of LFP activity in different conditions. The pink line represents data under HA conditions, the red line represents data under NA conditions, the blue line represents data under NU conditions, and the light blue line represents data under LU conditions.

The shaded area of the curve represents the SEM. The right figure in each panel shows the single-trial data in the NA upper and NU lower conditions.

The color represents the voltage. The spatial-temporal distribution of awareness-related ERP activities after the appearance of the grating. The color represents the standardized voltage difference see Methods.

Population results of early phase ERP response in the prefrontal cortex. Different lines represent different conditions, as shown in Panel A. The right table shows the number of recording sites for different patients. The lower right is the location of these sites. The dots with different colors represent different subjects, as shown in Figure 2D.

As proposed above, the different activities between NA and NU are regarded as awareness-related activities. We found 89 sites with awareness-related iERP activity in the PFC for all recording sites. Figure 3B shows the spatial-temporal dynamics of visual awareness-related activities in all recording sites during the ms period sampled at 8 time points that equally divided this period, for a clearer display, projected to the cortex, see the Supplementary Materials for the complete video.

Visual awareness-related activities clearly appeared in the PFC during ms, and these early activities were mainly concentrated in the middle lateral prefrontal cortex LPFC and then spread to other brain areas of the PFC.

We further calculated the start time of awareness-related activity, i. Figure 3C shows the spatial distribution of all recording sites DOT standardized and convenient for visual display , and it can be found that the DOT of the PFC is not explicitly later than those in posterior brain regions, including the posterior parietal lobe, temporal lobe, occipital lobe, etc.

We focused on the sites in PFC with earlier DOTs. We focused on the early DOT sites and calculated the normalized average iERP response of the early DOT sites Figure 3D.

At the population level, visual awareness-related activities in the PFC were also significant and started at approximately ms, which is located in the VAN time window Figure 3D , left.

That is, the early visual awareness-related activities in the prefrontal cortex are robust and consistent not only at the group level but also at the individual level. Although the above iERP results indicate that there are early visual awareness-related neural activities in the prefrontal cortex, it is difficult to explain whether this response is the result of local processing in the PFC or transmission from other brain regions.

Therefore, we calculated the LFP response in the high-gamma HG band, which is more representative of the local processing, of the prefrontal cortex. Figure 4A shows the exemplified high-gamma activity in the prefrontal cortex from each patient at the grand average left , spectrogram middle , and single-trial level right.

It is similar to the iERP results Figure 3 except the data are the HG power and there was a vigorous visual evoked power increase under the HA condition compared to the LU condition.

This difference is also very explicit and consistent at the single-trial and single-subject levels. The HG activity divergence onset time DOT of different recording sites in the PFC was also consistent within the early stage, which was similar to the above iERP result.

High-gamma activity of example sites from each patient. In each panel, the left figure shows the grand average of high-gamma power in different conditions. The color represents the normalized power. The spatial-temporal distribution of awareness-related high-gamma activities after the appearance of the grating.

The color represents the standardized power difference. Spatial distribution of divergence onset time for high-gamma activity. The color represents the normalized DOT value. Population results of early phase high-gamma response in the prefrontal cortex. The middle panel shows the spectrogram in NA upper and NU lower trials.

Percentage of awareness-related sites in ERP and HG analysis. n, number of recording sites in PFC. We found 31 sites with awareness-related HG activity in the PFC 74 for all recording sites.

Figure 4B shows the spatial-temporal dynamics of visual awareness-related HG activities in all recording sites during the ms period for a clearer display, projected to the cortex, see the Supplementary Materials for the complete video.

Visual awareness-related HG activities also clearly appeared in the PFC during ms, while they were rarely seen in lower-level visual areas occipital and temporal lobe , and these early activities were mainly concentrated in the middle lateral prefrontal cortex LPFC and then spread to other brain areas of the PFC.

We further calculated the start time of awareness-related HG activity, i. Figure 4C shows the spatial distribution of all recording sites DOT standardized and convenient for visual display , and it can be found that the DOT of the PFC is not explicitly later than those in posterior brain regions, and the earliest region is still located in the middle lateral prefrontal cortex.

We focused on the early DOT sites and calculated the normalized mean response Figure 4D and spectrum of the early DOT sites. Page last reviewed: 27 May Next review due: 27 May Home Health A to Z Back to Health A to Z.

Overview - Disorders of consciousness Contents Overview Causes Diagnosis. Consciousness requires both wakefulness and awareness. The main disorders of consciousness are: coma vegetative state minimally conscious state Coma A coma is when a person shows no signs of being awake and no signs of being aware.

Read more about comas. Vegetative state A vegetative state is when a person is awake but is showing no signs of awareness. A person in a vegetative state may: open their eyes wake up and fall asleep at regular intervals have basic reflexes such as blinking when they're startled by a loud noise or withdrawing their hand when it's squeezed hard They're also able to regulate their heartbeat and breathing without assistance.

They also show no signs of experiencing emotions. If a person is in a vegetative state for a long time, it may be considered to be: a continuing vegetative state when it's been longer than 4 weeks a permanent vegetative state when it's been more than 6 months if caused by a non-traumatic brain injury, or more than 12 months if caused by a traumatic brain injury If a person is diagnosed as being in a permanent vegetative state, recovery is extremely unlikely but not impossible.

Minimally conscious state A person who shows clear but minimal or inconsistent awareness is classified as being in a minimally conscious state.

But it's more difficult to diagnose a permanent minimally conscious state because it depends on things like: the type of brain injury how severe the injury is how responsive the person is In most cases, a minimally conscious state isn't usually considered to be permanent until it's lasted for several months and there have been no signs of improvement.

Why they happen Disorders of consciousness can occur if the parts of the brain involved with consciousness are damaged.

These types of brain injury can be divided into: traumatic brain injury — the result of a severe head injury , such as an injury sustained during a car accident or a fall from a great height non-traumatic brain injury — where the injury to the brain is caused by a health condition, such as a stroke progressive brain damage — where the brain is gradually damaged over time for example, because of Alzheimer's disease Read more about the causes of disorders of consciousness.

Making a diagnosis A disorder of consciousness will only be confirmed after extensive testing to determine the person's level of wakefulness and awareness. Treatment and care Treatment can't ensure recovery from a state of impaired consciousness. This can involve: providing nutrition through a feeding tube making sure the person is moved regularly so they don't develop pressure ulcers gently exercising their joints to prevent them becoming tight keeping their skin clean managing their bowel and bladder for example, using a tube known as a catheter to drain the bladder keeping their teeth and mouth clean offering opportunities for periods of meaningful activity — such as listening to music or watching television, being shown pictures or hearing family members talking Sensory stimulation In some cases, a treatment called sensory stimulation may be used in an attempt to increase responsiveness.

Some examples of sensory stimulation include: visual — showing photos of friends and family, or a favourite film hearing — talking or playing a favourite song smell — putting flowers in the room or spraying a favourite perfume touch — holding their hand or stroking their skin with different fabrics It's not entirely clear how effective sensory stimulation is, but it's sometimes considered worthwhile.

Recovery It's impossible to predict the chances of someone in a state of impaired consciousness improving. It largely depends on: the type of brain injury how severe the injury is the person's age how long they have been in the state for Some people improve gradually, whereas others stay in a state of impaired consciousness for years.

Withdrawing nutritional support If a person is in a permanent vegetative state for 12 months after a traumatic brain injury, or 6 months after a non-traumatic brain injury, then it may be recommended to withdraw their nutritional support.

This is because: there's almost no chance of a recovery by this point prolonging life would have no benefit for the person concerned prolonging treatment could offer false hope and cause unnecessary emotional distress for the friends and family of the person concerned The medical team will discuss the issue with family members.

Each of these topics relates to different states of consciousness because the way your mind processes your thoughts during each is unique. Human consciousness can be altered in a number of different ways, including through hypnosis, with drugs, and mental exercises.

Reaching an altered state of consciousness can also be achieved through various types of meditation and mind-body practices. A person's state of consciousness is interconnected with their level of awareness. Conversely, when exposed to a stimulant, a person can experience a heightened level of awareness.

Although you may not feel as if you are aware of every single detail of what is happening around you, even in a state of consciousness where you have a low level of awareness, your brain is still capable of processing the signals it receives. Think about the way a person who is asleep in bed may instinctively grab a blanket when they feel cold.

Although they are not actively thinking about being cold because of their current state of consciousness during sleep, their brain received a signal that their body feels cold.

A person who experiences a high level of awareness tends to be more in control of their thoughts. The ability to pay attention to detail and analyze the activity around you comes with heightened awareness, but this can typically only be achieved during certain states of consciousness.

The practice of mindfulness is one example of how heightened awareness of a person's thoughts can be achieved by focusing on the present moment. Meditation is often used to reach this type of awareness because the practice can help individuals achieve an altered state of consciousness.

States of consciousness play an important role in our everyday experiences. In order to see how consciousness works, it helps to understand how consciousness can be influenced by different factors.

For example, many people begin the day full of energy but then start to feel run-down by mid-afternoon. Other people struggle to plod through the morning routine only to finally feel energetic in the evening.

The daily fluctuations of energy levels are known as the circadian rhythm and play an important role in human consciousness.

Sometimes referred to as the body's "clocks," these daily rhythms have a major impact on your consciousness because they determine the level of awareness or alertness your body is able to achieve. Whether prescribed or illicit, chemical substances that impact a person's mental state can also affect their level of awareness.

Different types of drugs work by altering your state of consciousness in various ways. People can be in many different states of consciousness. Three states of consciousness to consider are sleeping, dreaming, and hypnosis.

Sleep has fascinated researchers, scientists, and scholars for thousands of years. Technological innovations during the last century allowed scientists to study sleep in ways that simply weren't possible in the past, which has helped lead to a number of theories of sleep.

Sleep creates a unique state of consciousness because while you experience a lowered level of awareness during sleep, your brain is still active.

Your brain's ability to cycle through different stages of sleep with both rapid eye movement REM and non-REM sleep is proof of this activity. When the sleep cycle is disrupted, it can wreak havoc on a person's body clock and impact their state of consciousness during both sleep and waking periods.

If you're not sleeping well , common sleep disorders could be to blame for the drowsiness, irritability, and brain fog that comes from disrupted sleep. REM sleep is the part of the sleep cycle most often associated with the dreams that you can remember when you wake up.

During this time, your brain is active in a way that is similar to when you are awake, but your state of awareness is lowered. At some point in your life, you have probably had a truly baffling dream that left you wondering, "What did that dream mean?

The phenomenon of hypnosis involves hyper-awareness which leads to an alteration of a person's state of consciousness. During hypnosis, a person will often appear as if they are asleep.

Although, in reality, they are actually processing their thoughts through a deep, focused awareness. Hypnosis has been used for a number of purposes including pain management and weight loss. Research also shows that hypnosis can be an effective treatment for different types of anxiety and pain.

There is no real consensus as to how many states of consciousness exist. This is, in part, because consciousness is multifaceted, and assessing levels of consciousness can be difficult.

Tononi G, Koch C. Consciousness: here, there and everywhere? Philos Trans R Soc Lond B Biol Sci. Kotchoubey B.

This notion of factorization is depicted in Fig. However, this factorization is asymmetric Fig. In the absence of awareness, there is by definition an absence of perceptual content, such that being unaware of a circle is similar to being unaware of a square.

This imposes an asymmetry in the model architecture which we will leverage in the next section when seeking to account for global ignition responses.

The model can be described formally in terms of a probabilistic graphical model Pearl, , where nodes correspond to unknown variables and the graph structure is used to indicate dependencies between variables. These graphs provide a concise description of how sensory data are generated Fig.

A Probabilistic graphical model of awareness reports. Nodes represent random variables and the graph structure is used to indicate dependencies as indicated by directed arrows. The shaded node indicates that this variable is observed by the system as sensory input.

B Expanded version of the graphical model from panel A that makes explicit the asymmetry in the state space. Figures created using the Daft package in Python. W is a 1 × N vector that encodes the relative probabilities of each of N discrete perceptual states. By expanding the graphical model to enumerate each discrete state Fig.

To simulate the model, I build on previous work using a 2D feature space to capture important features of multidimensional perceptual categorization King and Dehaene Each axis represents the strength of activation of one of two possible stimulus features, such as leftward or rightward tilted grating orientations see Fig.

The origin represents low activation on both features, consistent with no stimulus or noise being presented. As in the more general case described in the previous section, each stimulus category generates samples from a multivariate Gaussian whose mean is dominated by one or other feature.

A 2D feature space for a toy perceptual decision problem involving classifying two possible stimuli e. left- and right-tilted Gabors. Each Gaussian indicates the likelihood of observing a pair of features e. The right-tilted stimuli occupy the right-hand side of the grid; left-tilted stimuli occupy the left-hand side of the grid.

The absence of stimulation is represented by a distribution in which activation of each feature is low towards the origin. In 2D SDT, there are three stimulus classes organized in a flat non-hierarchical structure.

B The same 2D feature space from A , modified to make explicit the hierarchical aspect of the HOSS model. A higher-order awareness state a 1 nests perceptual states w 1 and w 2. King and Dehaene showed that by placing different types of decision criteria onto this space, multiple empirical relationships between discrimination performance, confidence and visibility could be simulated.

Our model builds closely on theirs and inherits the benefits of being able to accommodate dissociations between forced-choice responding and awareness reports. However, it differs in proposing that awareness is not inherent to perceptual categorization; instead, perceptual categorization is nested under a generative model of awareness Fig.

We will see that this seemingly minor change in architecture leads to important consequences for the relationship between awareness and global ignition. To explore the properties of the model, I simulate inference at different levels of the hierarchy for the two-class stimulus discrimination problem described in Fig.

Figure 4A shows that this probability rises in a graded manner from the lower left corner of the graph low activation of any feature to the upper right high activation of both features. In contrast, confidence in making a discrimination response e. rightward vs. leftward increases away from the major diagonal Fig.

In panel B , confidence in stimulus identify e. left- or right-tilted Gabor increases towards the corners of the grid. Overlaid in white is the 0. I next simulate a proxy for prediction error at each layer in the model—in other words, how much belief change was induced by the sensory sample.

I use the Kullback—Leibler K-L divergence as a compact summary of how far the posterior probability distribution at each level in the network differs from the prior. Flat priors were used for both the A and W levels. The K-L divergence is a measure of Bayesian surprise, which under predictive coding accounts is linked to neural activation at each level in a hierarchical network Friston ; Summerfield and de Lange For instance, if the model starts out with a strong prior that it will see gratings of either orientation, but a grating is omitted, this constitutes a large prediction error an unexpected absence.

This is due to the large belief updates invoked in the perceptual layer W by samples that deviate from the origin. In contrast, when we compute K-L divergence for the awareness state Fig.

This is because at this level inference on presence and absence is symmetric. A, B K-L divergence for A perceptual states W and B awareness state A as a function of sensory input X. K-L divergence quantifies the change from prior to posterior after seeing the stimulus X and provides a metric for the magnitude of belief update at different levels of the network.

The network nodes correspond to those in Fig. C Behaviour of the network in a simulated masking experiment at various levels of SOA modelled as increasing sensory precision in which sensory evidence was sampled from the three stimulus classes shown in Fig. The model is currently situated at a computational level and remains agnostic about temporal dynamics and neural implementation.

For recent work translating probabilistic graphical models into models of neuronal message passing see George et al.

Here, I instead focus on coarser-scale predictions about the neural correlates of awareness reports in typical consciousness experiments.

First, as hinted above, an asymmetric state space for presence and absence suggests there will be greater summed prediction error in the entire network on presence decisions as summarized by K-L divergence at each node of W.

This may be a computational correlate of the global ignition responses often found to track awareness reports Del Cul et al. Second, the model predicts that awareness reports but not discrimination performance, which relies on lower-order inference on W will depend on higher-order states. These may be instantiated in neural populations in prefrontal and parietal cortex Lau and Rosenthal Thus, it may be possible to silence or otherwise inactivate the neural substrates of an awareness state without affecting performance—a type of blindsight Weiskrantz ; Del Cul et al.

Third, for the uppermost awareness state, we expect symmetry—decisions in favour of both presence and absence will lead to belief updates of similar magnitude.

There has been limited focus on examining decisions about stimulus absence as these decisions are often used as a baseline or control condition in studies of perceptual awareness. However, existing data are compatible with symmetric encoding of presence and absence at the upper level of the visual hierarchy, in primate lateral prefrontal cortex LPFC; Panagiotaropoulos et al.

Merten and Nieder trained monkeys to report the presence or absence of a variety of low-contrast shapes presented near to visual threshold. Neural activity tracking the decision present or absent was decorrelated from that involved in planning a motor response by use of a post-stimulus cue that varied from trial to trial.

Importantly the magnitude of activation of these populations was similar in timing and strength, suggesting a symmetric encoding of awareness in LPFC. Using fMRI, Christensen et al. More broadly, the current framework suggests that focusing on inference about absence will be particularly fruitful for understanding the neural and computational basis of conscious awareness Merten and Nieder ; Farennikova ; Merten and Nieder The current simulations assume that the noise of sensory input is fixed, but in reality this parameter would also need to be estimated.

A range of disambiguating cues are likely to prove important in such an estimation scheme and thereby affect an inference on awareness. For instance, beliefs about the state of attention or other properties of the sensory system provide global, low-dimensional cues as to the state of awareness Lau ; Graziano Computationally these cues may be implemented as beliefs about precision priors on Σ in the model in Fig.

Another source of information about sensory precision is proprioceptive and interoceptive input about bodily states. Consider the following thought experiment in which we set up two conditions in a dark room, one in which the subject has their eyes open and one in which they have their eyes closed.

In other words, proprioception provides disambiguating information as to the current state of awareness—when the eyes are open we expect to have higher precision input than when the eyes are closed.

One straightforward way of introducing this relationship is to allow precision itself to depend on awareness a connection between A and Σ in Fig. Such a modification implies that an awareness state may be 2D, encoding the distinction between whether something has the potential to be seen high vs.

low expected precision or eyes open vs. eyes closed as well as whether something is seen present vs. absent; Metzinger ; Limanowski and Friston However, in contrast to the attention schema approach, in HOSS a model of attention provides a critical input into resolving ambiguity about whether we are aware or not by affecting beliefs about precision , rather than determining awareness itself.

The goal of the higher-order state-space HOSS approach outlined here is modest—to delineate computations supporting metacognitive reports about awareness. This is a useful place to start given that report or the potential for report is the jumping-off point for a scientific study of consciousness.

A stronger reading of the model is that conscious awareness and metacognitive reports depend on shared mechanisms in the human brain Shaver et al. Notably, a process of hierarchical inference may take place via passive message-passing without any strategic, cognitive access to this information e.

in working memory Carruthers , making it compatible with higher-order representational accounts of phenomenal consciousness Brown However, while inference on higher-order states is, on this view, necessary for awareness, it may not be sufficient.

In HOSS, the higher-order awareness state is simple and low-dimensional. However, it seems an empirical question as to the relative granularity of higher-order and first-order representations, and a range of intermediate views are plausible.

The more important point is that the state space is factorized to allow two separate causes of the sensory data—what it is, and whether I have seen it. This is likely to be computationally demanding in a rich, multidimensional state space as it requires marginalizing over W and may be rare in most cognitive systems.

HOSS also provides a new perspective on global workspace GWS architectures. As a result of global broadcast, cognitive and linguistic machinery have access to information about a particular stimulus or subpersonal mental state Dehaene et al.

Such a move potentially resolves the conundrum of how global broadcast directly accounts for systems claiming to be conscious of a stimulus without positing additional machinery.

This point is made concisely by Graziano Graziano :. It does not compute in this domain. In other words, it is difficult to see how such a system can be actively aware of the absence of stimulation when global broadcast is constitutive of awareness.

The state space approach outlined here is designed explicitly to compute in this domain, and therefore does not suffer from the same problem. Another critical difference between GWS and HOSS is that HOSS predicts prefrontal involvement for active decisions about stimulus absence , whereas GWS predicts that PFC remains quiescent on such trials due to a failure of the stimulus to gain access to the workspace.

Finally, to the extent that abstract awareness states need to be learnt or constructed, creating this level may require a protracted period of development. Such development would begin with creating a perceptual generative model W before a more general property awareness could be abstracted from these perceptual states.

I close with questions for future research motivated by the current computational sketch:. How are awareness states represented in neural activity? Are presence and absence encoded symmetrically?

Is a neural representation of awareness factorized with respect to other aspects of perceptual content? I am grateful to the Metacognition and Theoretical Neurobiology groups at the Wellcome Centre for Human Neuroimaging and members of the University of London Institute of Philosophy for helpful discussions.

I thank Matan Mazor, Oliver Hulme, Chris Frith, Nicholas Shea, Karl Friston and two anonymous reviewers for comments on previous drafts of this manuscript. Aru J , Bachmann T , Singer W , et al. Distilling the neural correlates of consciousness.

Neurosci Biobehav Rev ; 36 : — Google Scholar. Baars BJ. A Cognitive Theory of Consciousness. Cambridge, UK : Cambridge University Press , Google Preview. Block N.

On a confusion about a function of consciousness. Brain Behav Sci ; 18 : — Perceptual consciousness overflows cognitive access. Trends Cogn Sci ; 15 : — Brown R. The HOROR theory of phenomenal consciousness. Philos Stud ; : — Brown R , Lau H , Le Doux J. The Misunderstood Higher-Order Approach to Consciousness.

When that wave started to die down towards the late nineties, the next wave of innovations began with the introduction of a rush of innovation-enhancing PROCESSES.

Suddenly, there were numerous releases of innovation methodologies, and I would come across innovation consultants who could be called upon to facilitate innovation teams through innovation processes.

Processes like brainstorming sessions brought together a bunch of stakeholders, wildly brainstorming using post-its, clustering them and evaluating these ideas against a select set of criteria. Then they would instantly go into the lab and start quickly creating a very simple prototype.

Processes like Design Thinking and AGILE are very current and popular methodologies where inventors are made to step into the reality of the user, reenact the experience of the user, and use empathy and ideation to become part of the environment where a product is used.

When we listen into the industry, we hear that in spite of all these investment, innovation is drying out simply because process thinking is peaking. Companies are now desperately searching for the next wave of thinking for accelerating growth.

We at BEING AT FULL POTENTIAL recognize that the next wave of innovations is going to be unleashed by pulling the human lever. It is going to come from harnessing the human states — not systems and processes — by moving people into heightened states of awareness where they become infinitely creative.

In the first state, a state that we call the State of Primitive Awareness , our mental faculties are so blocked that we cannot see beyond the problem. Consumed by fear or anxiety we are unable to see solutions. You present a problem to a person who is in this state, and he just freezes.

When they are in such a fear-driven or anxious state, they cannot see beyond the problem. They do not see any solution to a problem.

No amount of reasoning or rational conversation is helpful when someone is in this state. When people are raised to a slightly higher state, a more expanded inner state, they can see two solutions to every problem.

We call this state the State of Binary Awareness. In this higher state of awareness, people see two solutions to every problem; Yes or No, Do or Die, Black or White, Left or Right, Stay or Leave, Democrat or Republican. Suppose you know someone who is having problems in her relationship.

If you ask her what solutions she sees, she will come up with two options; one is that she will simply continue with her relationship as it is, or the second solution that she would see is to say good bye to that relationship. In this state of awareness, innovation is binary. Most people I meet are in this state of awareness where they see two solutions to every problem.

There is a higher state, a third level of awareness. A person with locked-in syndrome is both conscious and aware, but completely paralysed and unable to speak.

They're usually able to move their eyes and are sometimes able to communicate by blinking. Page last reviewed: 27 May Next review due: 27 May Home Health A to Z Back to Health A to Z. Overview - Disorders of consciousness Contents Overview Causes Diagnosis.

Consciousness requires both wakefulness and awareness. The main disorders of consciousness are: coma vegetative state minimally conscious state Coma A coma is when a person shows no signs of being awake and no signs of being aware. Read more about comas. Vegetative state A vegetative state is when a person is awake but is showing no signs of awareness.

A person in a vegetative state may: open their eyes wake up and fall asleep at regular intervals have basic reflexes such as blinking when they're startled by a loud noise or withdrawing their hand when it's squeezed hard They're also able to regulate their heartbeat and breathing without assistance.

They also show no signs of experiencing emotions. If a person is in a vegetative state for a long time, it may be considered to be: a continuing vegetative state when it's been longer than 4 weeks a permanent vegetative state when it's been more than 6 months if caused by a non-traumatic brain injury, or more than 12 months if caused by a traumatic brain injury If a person is diagnosed as being in a permanent vegetative state, recovery is extremely unlikely but not impossible.

Minimally conscious state A person who shows clear but minimal or inconsistent awareness is classified as being in a minimally conscious state. But it's more difficult to diagnose a permanent minimally conscious state because it depends on things like: the type of brain injury how severe the injury is how responsive the person is In most cases, a minimally conscious state isn't usually considered to be permanent until it's lasted for several months and there have been no signs of improvement.

Why they happen Disorders of consciousness can occur if the parts of the brain involved with consciousness are damaged. These types of brain injury can be divided into: traumatic brain injury — the result of a severe head injury , such as an injury sustained during a car accident or a fall from a great height non-traumatic brain injury — where the injury to the brain is caused by a health condition, such as a stroke progressive brain damage — where the brain is gradually damaged over time for example, because of Alzheimer's disease Read more about the causes of disorders of consciousness.

Making a diagnosis A disorder of consciousness will only be confirmed after extensive testing to determine the person's level of wakefulness and awareness. Treatment and care Treatment can't ensure recovery from a state of impaired consciousness. This can involve: providing nutrition through a feeding tube making sure the person is moved regularly so they don't develop pressure ulcers gently exercising their joints to prevent them becoming tight keeping their skin clean managing their bowel and bladder for example, using a tube known as a catheter to drain the bladder keeping their teeth and mouth clean offering opportunities for periods of meaningful activity — such as listening to music or watching television, being shown pictures or hearing family members talking Sensory stimulation In some cases, a treatment called sensory stimulation may be used in an attempt to increase responsiveness.

Some examples of sensory stimulation include: visual — showing photos of friends and family, or a favourite film hearing — talking or playing a favourite song smell — putting flowers in the room or spraying a favourite perfume touch — holding their hand or stroking their skin with different fabrics It's not entirely clear how effective sensory stimulation is, but it's sometimes considered worthwhile.

Recovery It's impossible to predict the chances of someone in a state of impaired consciousness improving. It largely depends on: the type of brain injury how severe the injury is the person's age how long they have been in the state for Some people improve gradually, whereas others stay in a state of impaired consciousness for years.

The brain response to TMS perturbation was already used to distinguish the levels of awareness, irrespective of sensory processing and motor responses under physiological, pharmacological, and pathological conditions 34 , 35 , Therefore, this TMS-evoked response was applied to our end-to-end CNN framework.

Similar to several studies using ML, we observed higher two-class low or high classification accuracy using CNN when compared to linear discriminant analysis LDA and support vector machine SVM for both arousal and awareness.

This suggests that our framework is especially relevant for EEG results, which possess several nonlinear features. The classification performance of spatiotemporal information was higher than that of spatio-spectral information.

Using our framework, it was shown that temporal information discriminates different levels of consciousness more clearly than spectral information, as the functional connectivity associated with consciousness changes in both space and time However, this does not imply that temporal information is more important than spectral information for distinguishing consciousness.

Temporal information has more distinct characteristics than spectral information for predicting the state of consciousness in the proposed framework. Nevertheless, PCI also used spatiotemporal dynamics in TMS-evoked responses 16 , which is significantly important for distinguishing consciousness.

We applied transfer learning to a single domain as well as multiple domains. In the sleep domain, classification performance was high when trained on only the sleep domain or together with the anesthesia domain. Due to a large number of trials in the sleep domain, training was performed satisfactorily using only sleep data.

The distance of the averaged TMS-evoked potentials under the domains of anesthesia and sleep was significantly close, which indicated that the two domains were highly similar Thus, a close distance implies that the two domains have similar patterns, and the classification performance indeed increased when these domains were trained together.

In the anesthesia and DoC domains, the trials of a single domain were not sufficient; thus a higher classification performance was achieved when trained with similar domains. In addition, brain signals change over time, even when recorded with the same participants because of physiological and psychological differences over time Therefore, participant-independent learning such as the LOPO cross-validation is considerably difficult, as opposed to participant-dependent learning We solved these problems via transfer learning utilizing multiple domains and participant-independent ECI.

ECI, by averaging interclass probability, distinguished whether each state was low or high in both arousal and awareness. Because of its inherently high two-class performance, it was possible to calculate the discriminable ECI using a few trials in a single session.

The criteria used for determining ECI aro and ECI awa were different, especially for REM sleep, ketamine-induced anesthesia, and UWS patients. Although the dataset was the same, these states were trained by different labels depending on arousal and awareness.

For instance, REM sleep is high in awareness but low in arousal. The proposed classifier learns models for these different criteria by training itself based on the criteria of arousal and awareness.

Specifically, because the learning ability of the CNN is derived from the automatic extraction of complicated representations from EEG signals 39 , it can properly distinguish between both states, even if the same state has different labels depending on the criteria.

According to the LRP, we observed higher relevance scores in the parietal regions, compared to other regions at the scalp level. This was observed for all data, including TMS—EEG and resting-state EEG results.

The brain regions that led to decision-making were similar in arousal and awareness. However, this does not imply that arousal and awareness are supported by the same underlying neurophysiological mechanisms. The relevance scores simply explain the patterns of cortical EEG activity resulting in this classification.

In sleep and healthy wakefulness, when arousal and awareness were high, they were highly relevant in the parietal region. This EEG feature, which distinguishes high and low states in arousal and awareness, can be interpreted in line with the posterior hot zone of consciousness Local changes in this parietal region are associated with the occurrence of dreaming and unconscious sleep 41 , 42 , and our framework may learn from EEG pivotal features recorded in this area.

The importance of the posterior hot zone has already been emphasized using the within-state sleep paradigm 43 , Similar to sleep, we observed high relevance scores for the parietal region in the domains of both anesthesia and patients with DoC.

This implies that EEG activity in this region had a decisive effect in determining the high and low states of arousal and awareness. The increased slow-wave activity was observed under propofol- and xenon-induced anesthesia when compared to healthy wakefulness before drug administration In addition, just as cortical neurons induced bistable changes during NREM sleep, TMS in propofol-mediated anesthesia-induced low-amplitude, low-frequency, positive—negative potentials, and TMS in xenon-mediated anesthesia caused a significantly large amplitude but stereotyped positive—negative deflection Moreover, under ketamine-induced anesthesia, TMS-evoked response was determined to be similar to REM sleep, which features dreaming during a low state of arousal Previous studies have shown that the change in slow waves induced by propofol is primarily observed in the posterior hot zone 45 and the posterior main hub is disrupted during anesthesia-induced alteration of consciousness 7.

In UWS patients, TMS triggered a local and slow response similar to NREM sleep and general anesthesia, whereas MCS patients showed complex TMS-evoked responses Similarly, differences in alpha connectivity between the UWS and MCS patients are apparent within the posterior hot zone That is, when determining whether arousal and awareness are low or high, our classifiers used differences and changes in EEG activity over the parietal region to make decisions.

Particularly, similar results were observed in patients with DoC using several TMS target sites, and in sleep and anesthesia domains, where the parietal region was primarily stimulated.

This suggests that our findings regarding the parietal region are unrelated to the TMS target site. Thus, our trained model used neurophysiological features to classify whether arousal and awareness are low or high.

This indicates an appropriate design of our model as the classification decision was primarily based on EEG signals over the parietal region, which is suggested to be a hot spot of consciousness, compared to the frontal region The difference in the parietal region could be clearly identified through correct trials during sleep and healthy wakefulness.

It is meaningful that the frontal region contributed less than parietal regions in the context of the controversy regarding the spatial localization of the neural correlates of consciousness 40 , Considering the subcortical influences related to striatal-thalamic circuits, it has been recently observed that the parietal region contributes more to the levels of consciousness than the frontal region The implication of the parietal cortex in consciousness has also been demonstrated in other neuroimaging modalities, such as functional magnetic resonance imaging 48 and magnetoencephalography This study does have certain limitations.

First, our sample size was relatively small. In the future, it will be necessary to further test the reliability of the proposed indicator with larger cohorts and validate it at the clinical level before implementing it in a clinical setting. Sleep experiments would also have to be applied to more participants in the future.

Second, we explored the possibility of calculating ECI with a minimal amount of data, up to a single trial. However, we did not attempt to measure ECI in real time. Thus, in the future, ECI could be calculated in real-time for practical application.

Third, ECI does not differentiate between physiological, pharmacological, and pathological conditions, but distinguishes between high and low states of arousal and awareness. ECI can thus distinguish between REM sleep or ketamine and wakefulness. It may also be difficult to select the model to use when calculating ECI since the domain has to be known beforehand.

Nevertheless, if a single domain has sufficient trials, the LOPO approach would be the most accurate. Another limitation might be related to the possible contamination of TMS-EEG data by auditory and somatosensory components. As in previous studies 17 , 35 , 36 , to avoid auditory and somatosensory co-stimulation, participants wore earphones with noise masking and a thin foam between the scalp and the TMS coil that was used.

Although it is difficult to systematically rule out the contribution of sensory co-stimulation in every measurement, the application of effective noise-masking procedures 50 and the real-time monitoring of data quality during the acquisition 51 may significantly mitigate this issue Finally, ECI indicates whether arousal and awareness are low or high and cannot be considered functional.

It, therefore, should be developed into a functional index. In conclusion, we proposed ECI as a neurophysiological indicator to simultaneously discriminate the levels of arousal and awareness in modified states of consciousness.

This tool allows disentangling the levels of consciousness, with a single measure, in different clinical settings such as monitoring surgical interventions i. This indicator was validated under different physiological, pharmacological, and pathological conditions, and it reliably disentangled low levels from high levels of both arousal and awareness.

Besides, the proposed ECI is considerably accessible and practical, as it can be applied to resting-state EEG without TMS, and requires fewer trials. Therefore, the proposed indicator can be a reliable discriminator and valuable tool as an objective measure of consciousness. As parietal regions appear to be the most relevant for classification, an EEG configuration around that area could be sufficient if ECI is used in clinical practice.

These findings could be useful in diagnosing severely brain-injured patients and monitoring their levels of consciousness in real-time, especially in clinical settings where time constraints preclude long-duration assessment. The proposed reliable ECI can provide insights into the classification of conscious levels using deep learning and neural correlates of consciousness.

The sleep dataset included six healthy participants five males, aged The inclusion criteria included i between 18 and 75 years of age and ii in good general health. The exclusion criteria were as follows: i neurological, psychiatric, mood, and sleep disorders, ii contraindications for TMS e.

All participants provided written consent, and the experimental paradigm was approved by the Institutional Review Board IRB at the University of Wisconsin—Madison HSC Supplementary Table 8 lists the TMS target site and the number of sessions and trials. The participants were awoken by an alarm sound that lasted for 1.

They were then asked if they had had conscious experience. The TMS—EEG experiments were performed over a period of four or five nights per participant. The anesthesia data were previously published by Sarasso et al. The inclusion criteria included i older than 18 years and ii stability of vital parameters.

The exclusion criteria were as follows: i neurological, cardiovascular, psychiatric, and mood disorders, ii contraindications for TMS e. TMS was applied over the left parietal or motor regions after participants reached deep unresponsiveness a score equal to 5 in the Ramsay scale, which corresponds to no response to external stimuli following standard anesthetic procedures The stimuli target site and the number of trials are listed in Supplementary Table 9.

In addition, upon waking up from anesthesia, reports about the conscious experience during anesthesia were collected. Conclusively, the participants reported little conscious experience during propofol- and xenon-induced anesthesia, but vivid dreams were experienced during ketamine-induced anesthesia.

For patients with severe brain injury, the data of six UWS patients 2 males, 4 traumatic brain injuries, time since the injury of All of them fell into a coma due to brain injury and presented a prolonged state of impaired consciousness.

The inclusion criteria included i older than 18 years and ii diagnosis of DoC following a severe acquired brain injury. The exclusion criteria for patients were as follows: i patients having significant neurological, neurosurgical, or psychiatric disorders prior to the brain injury that leads to DoC, ii patients having any contraindication to TMS—EEG or magnetic resonance imaging electronic implanted devices, active epilepsy, external ventricular drain , and iii patients who were not medically stable.

Accredited experts performed repeated CRS-R for each patient, including on the day of the TMS—EEG examination and before the fluorodeoxyglucose-positron emission tomography FDG-PET scan. TMS—EEG data were acquired similar to previous studies 16 , Moreover, added data were newly included as follows: 9 UWS patients 6 males, 5 traumatic brain injuries, time since the injury of 6.

These added data were recently acquired by Dr. Steven Laureys team at the University of Liège. For the added data, the inclusion and exclusion criteria for patients were exactly the same as the previously used data. These data are part of a bigger study conducted in the frame of the Human Brain Project.

The detailed demographic and clinical information of severely brain-injured patients is listed in Supplementary Table The TMS target site was selected using a neuronavigation system over the parietal, motor, or premotor regions, avoiding structural lesions using the magnetic resonance imaging data of the patient.

The stimuli target sites for all participants are listed in Supplementary Table The participants remained awake or were kept awake using the CRS-R arousal protocol in between TMS stimulation During all the sessions, earphones presenting white noise were used to reduce the noise of the TMS pulses and we used a thin foam between the scalp and the TMS coil to avoid somatosensory evoked potentials.

These pulses were presented at random intervals of 2—2. In particular, the TMS stimulation was always performed in the medial half of one hemisphere, to avoid any muscle artifacts. Finally, we used resting-state EEG without TMS for the domains of anesthesia and severely brain-injured patients.

These data were acquired using the same participants as in the TMS—EEG experiments. Finally, among previously published data, the data with a signal-to-noise of 1. TMS—EEG data were preprocessed using the SiSyPhus Project MATLAB program University of Milan, Italy and the EEGLAB toolbox The signals were down-sampled to Bad channels were manually detected and interpolated using superfast spherical interpolation for artifact removal.

The data were re-referenced to an average reference 58 , Resting-state EEG data were processed using the EEGLAB toolbox The number of trials for each session we used is listed for patients under anesthesia Supplementary Table 12 and severely brain-injured patients Supplementary Table More details related to this deliberate choice are reported in Supplementary Note 2.

In the first step, EEG data were converted from 2D raw signals to 3D input. To preserve the spatial information and characteristics of EEG, we used spatio-spectral and spatiotemporal 3D features. n indicates the number of electrode channels. However, these simple signals do not capture all the spatial information characteristics in the brain.

Therefore, we converted 1D data vectors to 2D EEG data meshes using the spatial information of the electrode location.

Zero was inserted in the place of a null electrode in 2D matrices at time index t Finally, we calculated 3D data by adding spectral or temporal information Supplementary Fig.

Specifically, spectral information was divided into 5 frequency bands: delta 1. Step 2—Calculation of interclass probability using CNN: The model was trained to distinguish both arousal and awareness in terms of whether they were low or high.

We first calculated the domain similarity based on the cosine distance for domain transfer learning. The cosine distance D between domains A and B is defined as follows:.

This similarity was used when selecting the source domain for domain transfer learning In calculating the similarity according to the states of arousal and awareness, the labels in each state are different.

During sleep, NREM sleep with no subjective experience and REM sleep with subjective experience were learned as low arousal, whereas healthy wakefulness was learned as high arousal. Conversely, in awareness, REM sleep with subjective experience and healthy wakefulness indicated by open eyes were learned as high awareness, while NREM sleep with no subjective experience was learned as low awareness.

Specifically, we used stage 3 of NREM sleep for clear feature extraction of deep sleep. Under general anesthesia and in healthy wakefulness before anesthesia, the arousal state was divided as follows: i low, when under ketamine, propofol, and xenon-induced anesthesia, and ii high, when in healthy wakefulness before anesthesia.

Conversely, the awareness state was divided into low under propofol and xenon-induced anesthesia and high under ketamine-induced anesthesia and healthy wakefulness before anesthesia.

Finally, in patients with DoC, UWS and MCS patients were distinguished in terms of awareness: low UWS patients and high MCS patients. The UWS and MCS patients corresponded to high arousal. Deep learning was conducted in a MATLAB environment powered by a TITAN V GPU. We used the LRP toolbox 30 for CNN classification and interpretation.

The CNN was applied to the two components of consciousness arousal and awareness. In each architecture, we inserted five convolutional layers with 2D filters for the deep neural network. Finally, the generated feature maps were flattened into a 1D vector.

A softmax layer was used for classification. In the softmax layer, each element indicates the probability that the original input belongs to the corresponding class.

In this training procedure, the parameters in the deep neural network were learned through back-propagation. The activation function in each convolutional layer was a rectified linear unit. The detailed CNN architecture is presented in Supplementary Table The Adam optimizer was used with an initial learning rate of 0.

Specifically, we used hyperparameters as values of 0. The batch size was 25 for training, and the maximum number of training iterations was five times the number of training data. Consequently, the output of this architecture was the probability of each class.

For a comparison with other classifiers, we also considered an LDA classifier 62 and SVM with polynomial kernel function 63 using the same input data as fair baseline methods.

The classification performance was measured in LOPO-nested cross-validation for the generalized neural network. This method is a special case of k -fold cross-validation. Specifically, all participants except one the target participant were used for training, and the target participant was then tested using the classifier.

This process was repeated for each participant. Further, internal validation sets inner cross-validation were performed to choose the hyperparameters of the model Thus, the same hyperparameters were selected in the external validation sets The LOPO cross-validation procedure uses data efficiently and can reduce overfitting.

It also provides unbiased estimates of the averaged classification error for all possible training sets The same LOPO-nested cross-validation is applied in LDA and SVM using the Berlin brain—computer interface toolbox Three possible values 0.

For spatio-spectral input and spatiotemporal input 1 and 0. Step 3—Calculation of ECI in a single session: We obtained each interclass probability according to two components arousal and awareness at the previous step. In each TMS session, the interclass probability was averaged to calculate an ECI.

This approach has the advantage of being able to offset outliers. N is the number of trials in a single session j. Consequently, ECI is expressed as a 2D indicator representing both arousal and awareness simultaneously. As mentioned earlier, we used the LOPO cross-validation.

It is to be noted that only data from one participant were used as the test, and the data from the remaining participants were used for training. Finally, the test and training data did not demonstrate an overlap at all.

Step 4—Interpretation using LRP: We used LRP based on a backward-propagation mechanism for interpretability of the deep neural networks. This calculated the pixel-based decomposition process This score is calculated through the backward propagation of the model input.

Therefore, relevance scores describe a single nonlinear decision for the output corresponding to each input Through this method, we can observe not only the interpretation of classification decisions but also what features the model has learned To investigate which brain regions and signals caused these classification results, we compared the relevance scores from the LRP by dividing them into the following three regions 69 : the frontal Fp1—2, Fpz, AF1—2, AFz, F1—8, and Fz , temporal FT9—10, T7—8, TP9—10 , and parietal CP1—6, CPz, P1—4, P7—8, and Pz regions.

We focused on the frontal, temporal, and parietal regions when comparing the relevance score resulting from the LRP as there is an ongoing debate regarding which brain area, i. We also included the temporal region as activation in the NREM sleep increases in this region Additional EEG analyses are shown in Supplementary Methods as follows: i performance according to the number of trials in the ECI calculation, ii comparison of the difference between correct and incorrect trials, and iii classification performance using EEG signals excluding frontal or parietal electrodes.

We compared ECI awa with the PCI values computed following the same procedure described in PCI measures the complexity of the spatiotemporal patterns of cortical activity significantly evoked by TMS PCI ranges between 0 minimum complexity and 1 maximum complexity.

The Kruskal—Wallis test was also performed to compare the classification performance using transfer learning. To investigate the discrimination of ECI in each state of consciousness, the feedforward network was trained with 20 hidden layers using the LOPO approach.

For each output class, the AUC, sensitivity, and specificity were calculated using ROC analysis. The Kruskal—Wallis test was employed to investigate if there were any differences in relevance scores among brain regions from LRP under sleep.

In addition, the Kruskal—Wallis test was performed to explore the differences in relevance scores from LRP under the condition of anesthesia and for severely brain-injured patients. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

All data sleep dataset, anesthesia dataset, already published brain injury dataset, and new added brain injury dataset generated and used in this study have been deposited in a local database and are available upon reasonable request to Olivia Gosseries.

The sleep dataset was previously reported by Nieminen et al. The anesthesia data were also previously published by Sarasso et al. The published brain injury dataset was previously reported by Bodart et al.

However, new added brain injury dataset has not yet been published. The raw EEG data are protected and are not made publically available owing to data privacy laws, but are available from the corresponding author upon reasonable request.

Source data are provided with this paper. Sanders, R. Unresponsiveness unconsciousness. Anesthesiology , — Article PubMed Google Scholar. Darracq, M. et al. Evoked alpha power is reduced in disconnected consciousness during sleep and anesthesia.

Article ADS PubMed PubMed Central Google Scholar. Colombo, M. The spectral exponent of the resting EEG indexes the presence of consciousness during unresponsiveness induced by propofol, xenon, and ketamine. Neuroimage , — Article CAS PubMed Google Scholar. Lendner, J. An electrophysiological marker of arousal level in humans.

eLife 9 , e Article CAS PubMed PubMed Central Google Scholar. Mashour, G. Neural correlates of unconsciousness in large-scale brain networks. Trends Neurosci. Casarotto, S. Exploring the neurophysiological correlates of loss and recovery of consciousness: perturbational complexity in Brain Function and Responsiveness in Disorders of Consciousness ed Monti, M.

Bonhomme, V. General anesthesia: a probe to explore consciousness. Article PubMed PubMed Central Google Scholar.

Incidence of connected consciousness after tracheal intubation: a prospective, international, multicenter cohort study of the isolated forearm technique. Noirhomme, Q. Giacino, J. The minimally conscious state: definition and diagnostic criteria.

Google Scholar. Gosseries, O. Measuring consciousness in severely damaged brains. The JFK Coma Recovery Scale-Revised: measurement characteristics and diagnostic utility.

Thibaut, A. Preservation of Brain Activity in Unresponsive Patients Identifies MCS Star. Recent advances in disorders of consciousness: focus on the diagnosis.

Brain Inj. Stender, J. Diagnostic precision of PET imaging and functional MRI in disorders of consciousness: a clinical validation study. Lancet , — Casali, A. A theoretically based index of consciousness independent of sensory processing and behavior. Stratification of unresponsive patients by an independently validated index of brain complexity.

PubMed PubMed Central Google Scholar. Engemann, D. Robust EEG-based cross-site and cross-protocol classification of states of consciousness. Brain , — Baird, B. Human rapid eye movement sleep shows local increases in low-frequency oscillations and global decreases in high-frequency oscillations compared to resting wakefulness.

eNeuro 5 , 4 Article Google Scholar. Müller, K. Machine learning for real-time single-trial EEG-analysis: from brain—computer interfacing to mental state monitoring.

Methods , 82—90 Lemm, S. Introduction to machine learning for brain imaging. Neuroimage 56 , — Liu, Q. IEEE Access 7 , — Fahimi, F. Inter-subject transfer learning with end-to-end deep convolutional neural network for EEG-based BCI.

Neural Eng. Article ADS PubMed Google Scholar. Webb, S. Deep learning for biology. Nature , — Montavon, G. Layer-wise relevance propagation: an overview in explainable AI: interpreting, explaining and visualizing deep learning eds Samek, W.

Kwon, O. Subject-independent brain-computer interfaces based on deep convolutional neural networks. IEEE Trans. Neural Netw. Sturm, I. Interpretable deep neural networks for single-trial EEG classification. Methods , — Lotte, F. A review of classification algorithms for EEG-based brain—computer interfaces: a 10 year update.

Article ADS CAS PubMed Google Scholar. Lapuschkin, S. The LRP toolbox for artificial neural networks. MathSciNet MATH Google Scholar. Massimini, M. Triggering sleep slow waves by transcranial magnetic stimulation. USA , — Article ADS CAS PubMed PubMed Central Google Scholar.

Slow waves, synaptic plasticity and information processing: insights from transcranial magnetic stimulation and high-density EEG experiments.

Napolitani, M. Transcranial magnetic stimulation combined with high-density EEG in altered states of consciousness. Breakdown of cortical effective connectivity during sleep. Science , — Rosanova, M. Recovery of cortical effective connectivity and recovery of consciousness in vegetative patients.

Sarasso, S. Consciousness and complexity during unresponsiveness induced by propofol, xenon, and ketamine. Luppi, A. Consciousness-specific dynamic interactions of brain integration and functional diversity.

Jeon, E. Domain adaptation with source selection for motor-imagery based BCI. in 7th International Winter Conference on Brain-Computer Interface BCI. Lawhern, V. EEGNet: a compact convolutional neural network for EEG-based brain—computer interfaces.

Koch, C. Posterior and anterior cortex—where is the difference that makes the difference? Siclari, F. Local aspects of sleep and wakefulness. Dreaming in NREM sleep: a high-density EEG study of slow waves and spindles.

Nieminen, J. Consciousness and cortical responsiveness: a within-state study during non-rapid eye movement sleep. Lee, M. Connectivity differences between consciousness and unconsciousness in non-rapid eye movement sleep: a TMS—EEG study. Network properties in transitions of consciousness during propofol-induced sedation.

Chennu, S. Brain networks predict metabolism, diagnosis and prognosis at the bedside in disorders of consciousness. Afrasiabi, M. Consciousness depends on integration between parietal cortex, striatum, and thalamus. Cell Syst. Vanhaudenhuyse, A. Default network connectivity reflects the level of consciousness in non-communicative brain-damaged patients.

Andersen, L. Cortex 26 , — Russo, S. TAAC—TMS Adaptable Auditory Control: a universal tool to mask TMS clicks. The rt-TEP tool: real-time visualization of TMS-evoked potentials to maximize cortical activation and minimize artifacts.

Belardinelli, P. Reproducibility in TMS—EEG studies: a call for data sharing, standard procedures and effective experimental control. Brain Stimul. Graph theoretical analysis of cortical networks based on conscious experience.

in 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society EMBC. Bodart, O. Measures of metabolism and complexity in the brain of patients with disorders of consciousness.

Neuroimage Clin. Global structural integrity and effective connectivity in patients with disorders of consciousness. Sleep-like cortical OFF-periods disrupt causality and complexity in the brain of unresponsive wakefulness syndrome patients.

Delorme, A. EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. Methods , 9—21 Bertrand, O. A theoretical justification of the average reference in topographic evoked potential studies.

Ludwig, K. Using a common average reference to improve cortical neuron recordings from microelectrode arrays. Zhang, D. Cascade and parallel convolutional recurrent neural networks on EEG-based intention recognition for brain computer interface.

in 32nd AAAI Conference on Artificial Intelligence, AAAI Kingma, D. Adam: A method for stochastic optimization. Blankertz, B. Single-trial analysis and classification of ERP components—a tutorial. Smits, G. Improved SVM regression using mixtures of kernels. in International Joint Conference on Neural Networks.

IJCNN'02 Cat. Rueda-Delgado, L. Brain event-related potentials predict individual differences in inhibitory control. Korjus, K. An efficient data partitioning to improve classification performance while keeping parameters interpretable. PLoS ONE 11 , e Thiery, T. Long-range temporal correlations in the brain distinguish conscious wakefulness from induced unconsciousness.

Neuroimage , 30—39 Trends Cogn Sci ; 15 : — Brown R. The HOROR theory of phenomenal consciousness. Philos Stud ; : — Brown R , Lau H , Le Doux J. The Misunderstood Higher-Order Approach to Consciousness. preprint, PsyArXiv, Carruthers P. Pac Philos Q ; 98 : 65 — Christensen MS , Ramsøy TZ , Lund TE , et al.

An fmri study of the neural correlates of graded visual perception. Neuroimage ; 31 : — Cleeremans A. The radical plasticity thesis: how the brain learns to be conscious. Front Psychol ; 2 : 1 — Dehaene S , Changeux J-P. Experimental and theoretical approaches to conscious processing.

Neuron ; 70 : — Dehaene S , Kerszberg M , Changeux JP. A neuronal model of a global workspace in effortful cognitive tasks. Proc Natl Acad Sci USA ; 95 : — Dehaene S , Naccache L , Cohen L , et al.

Cerebral mechanisms of word masking and unconscious repetition priming. Nat Neurosci ; 4 : Del Cul A , Baillet S , Dehaene S. Brain dynamics underlying the nonlinear threshold for access to consciousness. PLoS Biol ; 5 : e Del Cul A , Dehaene S , Reyes P , et al.

Causal role of prefrontal cortex in the threshold for access to consciousness. Brain ; : — Dennett DC. Consciousness Explained. Penguin UK , Green DM , Swets JA.

Signal Detection Theory and Psychophysics. New York : Wiley , Duncan J. The multiple-demand MD system of the primate brain: mental programs for intelligent behaviour.

Trends Cogn Sci ; 14 : — 9. Farennikova A. Seeing Absence. Philosophical Studies ; : — Fleming SM , Daw ND. Self-evaluation of decision-making: a general Bayesian framework for metacognitive computation.

Psychol Rev ; : 91 — Friston K. A theory of cortical responses. Philos Trans R Soc Lond B Biol Sci ; : — Friston KJ , Parr T , de Vries B. The graphical brain: Belief propagation and active inference. Netw Neurosci ; 1 : — Frith C , Perry R , Lumer E. The neural correlates of conscious experience: an experimental framework.

Trends Cogn Sci ; 3 : — George D , Lehrach W , Kansky K , et al. Science ; Graziano MSA. Consciousness engineered. J Conscious Stud ; 23 : 98 — Consciousness and the Social Brain. Oxford University Press , Hohwy J.

Prediction error minimization, mental and developmental disorder, and statistical theories of consciousness. In: Gennaro RJ ed. Cambridge, MA : MIT Press , , — The Predictive Mind. Kanai R , Walsh V , Tseng C.

Subjective discriminability of invisibility: a framework for distinguishing perceptual and attentional failures of awareness. Conscious Cogn ; 19 : — Kersten D , Mamassian P , Yuille A. Object perception as Bayesian inference.

Annu Rev Psychol ; 55 : — Kim C-Y , Blake R. Trends Cogn Sci ; 9 : — 8. King J-R , Dehaene S. A model of subjective report and objective discrimination as categorical decisions in a vast representational space.

Philos Trans R Soc Lond B Biol Sci ; : Lau H , Brown R. The empirical case for conscious experience without first-order representations. In: Pautz A , Stoljar D eds , Blockheads! Cambridge, MA : MIT Press , Lau H , Rosenthal D. Empirical support for higher-order theories of conscious awareness.

Lau HC , Passingham RE. Relative blindsight in normal observers and the neural correlate of visual consciousness. Proc Natl Acad Sci USA ; : — 8.

Lau HC. A higher order Bayesian decision theory of consciousness. Progr Brain Res ; : 35 — Limanowski J , Friston K. Front Psychol ; 9. Mante V , Sussillo D , Shenoy KV , et al. Context-dependent computation by recurrent dynamics in prefrontal cortex.

Nature ; : Martin J-R , Dokic J. Seeing absence or absence of seeing? Thought J Philos ; 2 : — Merten K , Nieder A. Active encoding of decisions about stimulus absence in primate prefrontal cortex neurons.

Proc Natl Acad Sci USA ; : — Comparison of abstract decision encoding in the monkey prefrontal cortex, the presupplementary, and cingulate motor areas. J Neurophysiol ; : 19 — Metzinger T. How does the brain encode epistemic reliability? Perceptual presence, phenomenal transparency, and counterfactual richness.

Cogn Neurosci ; 5 : — 4. Michel M , Morales M. Minority reports: consciousness and the prefrontal cortex. Mind Lang Panagiotaropoulos TI , Deco G , Kapoor V , et al. Neuronal discharges and gamma oscillations explicitly reflect visual consciousness in the lateral prefrontal cortex.

Neuron ; 74 : — Pearl J. Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference. Burlington, MA : Morgan Kaufmann Publishers , Ramsøy TZ , Overgaard M. Introspection and subliminal perception. Phenomenol Cogn Sci ; 3 : 1 — Sergent C , Dehaene S.

Is consciousness a gradual phenomenon? Evidence for an all-or-none bifurcation during the attentional blink. Psychol Sci ; 15 : — 8.

Shaver E , Maniscalco B , Lau HC. Awareness as confidence. Anthropol Philos ; 9 : 58 — Summerfield C , de Lange FP. Expectation in perceptual decision making: neural and computational mechanisms.

Nat Rev Neurosci ; 15 : — Tsuchiya N , Wilke M , Frässle S , et al. No-report paradigms: extracting the true neural correlates of consciousness.

Trends Cogn Sci ; 19 : — von Helmholtz H. Treatise on Physiological Optics Weiskrantz L. Consciousness Lost and Found: A Neuropsychological Exploration. Oxford : OUP , Oxford University Press is a department of the University of Oxford.

It furthers the University's objective of excellence in research, scholarship, and education by publishing worldwide.

Sign In or Create an Account. Navbar Search Filter Neuroscience of Consciousness This issue Consciousness Books Journals Oxford Academic Mobile Enter search term Search. Issues Volume , Issue 1, In Progress Volume , Issue 1, Browse all Special Issues Submit Author Guidelines Journal Policies Submission Site Open Access Why publish with this journal?

Call For Papers Alerts About About Neuroscience of Consciousness About the Association for the Scientific Study of Consciousness Editorial Board Advertising and Corporate Services Journals Career Network Self-Archiving Policy Journals on Oxford Academic Books on Oxford Academic.

Call For Papers Alerts About About Neuroscience of Consciousness About the Association for the Scientific Study of Consciousness Editorial Board Advertising and Corporate Services Journals Career Network Self-Archiving Policy Close Navbar Search Filter Neuroscience of Consciousness This issue Consciousness Books Journals Oxford Academic Enter search term Search.

Advanced Search. Search Menu. Article Navigation. Close mobile search navigation Article Navigation. Volume Article Contents Abstract. Psychological Basis of Awareness Reports. Empirical Predictions. Role of Precision Estimation in Resolving Awareness States.

Relationship to Other Theories of Consciousness. Research Questions. Data availability. Journal Article. Awareness as inference in a higher-order state space. Stephen M Fleming Stephen M Fleming. Wellcome Centre for Human Neuroimaging, Institute of Neurology.

Correspondence address. Wellcome Centre for Human Neuroimaging, University College London, 12 Queen Square, London WC1N 3AR, UK. fleming ucl. Oxford Academic. Revision received:. PDF Split View Views. Select Format Select format. ris Mendeley, Papers, Zotero.

enw EndNote. bibtex BibTex. txt Medlars, RefWorks Download citation. Permissions Icon Permissions. Close Navbar Search Filter Neuroscience of Consciousness This issue Consciousness Books Journals Oxford Academic Enter search term Search.

Figure 1. Open in new tab Download slide. Figure 2. Figure 3. Figure 4. Figure 5. Google Scholar Crossref. Search ADS. Google Scholar Google Preview OpenURL Placeholder Text.

Del Cul. Google Scholar OpenURL Placeholder Text.