CN113712507A

CN113712507A - System for evaluating degree of disturbance of consciousness, restoration tendency prediction method, and storage medium

Info

Publication number: CN113712507A
Application number: CN202110572592.4A
Authority: CN
Inventors: 王立平; 桂鹏; 蒋雨薇
Original assignee: Center for Excellence in Brain Science and Intelligence Technology Chinese Academy of Sciences
Current assignee: Center for Excellence in Brain Science and Intelligence Technology Chinese Academy of Sciences
Priority date: 2020-05-25
Filing date: 2021-05-25
Publication date: 2021-11-30
Anticipated expiration: 2041-05-25
Also published as: CN113712507B; WO2021237429A1

Abstract

The invention provides a system for evaluating the degree of disturbance of consciousness, a recovery tendency prediction method and a storage medium. The system comprises: a language generator configured to generate speech material that produces auditory stimuli to a subject; an electroencephalograph configured to record an electroencephalogram signal of the subject; and the analysis device is configured to extract characteristic parameters from the electroencephalogram signals and obtain the consciousness disturbance degree of the subject according to the characteristic parameters. The system can evaluate the degree of the disturbance of consciousness by adopting the bedside brain electricity, is convenient and fast to use, has accurate results, and can effectively predict the recovery tendency of the patient with disturbance of consciousness.

Description

System for evaluating degree of disturbance of consciousness, restoration tendency prediction method, and storage medium

Technical Field

The present invention relates generally to the field of cognitive neuroscience, and more particularly, to a system for evaluating the degree of an conscious disturbance, a recovery tendency prediction method, and a storage medium.

Background

Exploring residual consciousness and cognitive abilities in non-responsive patients is one of the important clinical concerns and a major challenge in cognitive neuroscience. Each year, there are thousands of new patients who lose communication due to brain damage and fall into different states of clinical consciousness, ranging from Coma (Coma) to Unresponsive Wakefulness Syndrome (UWS), and micro-consciousness State (MCS). The judgment of the degree of disturbance of consciousness of the state of a patient is mainly based on the evaluation of nonreflective behaviors such as movement, speech movement and the like at the bedside. At the moment when unresponsive arousal syndrome patients exhibit arousal, they may open their eyes and make complex reflex actions, but there are no signs that they exhibit clear intentional behavior. In contrast, patients with micro-conscious states exhibit some intentional behavior, but do not appear to establish a functional communication that lasts for a long period of time. Subtle differences between these two types of patients, often accompanied by other impaired performance, may lead to higher rates of misdiagnosis.

Recent studies have shown that several dynamic features of the electroencephalogram (EEG) signal may suggest levels of consciousness, such as the magnitude and latency of auditory evoked potentials, spectral energy, and signal complexity and functional connectivity via weighted sign mutual information. The mainstream theory of consciousness advocates that consciousness is a real-time evolution process of dynamic change, self-maintenance and whole brain cooperative work, rather than a static brain function. Accordingly, a series of recent studies report that the dynamic characteristics of functional magnetic resonance imaging (fMRI) in the resting state may provide a specific cortical indicator of unconsciousness. In particular, brain activity in the unconscious state is largely confined to dynamic patterns dominated by structural associations. In contrast, conscious states can be identified by more complex patterns of brain activity with long distance (e.g., frontal-parietal lobe) interactions. Similar to functional magnetic resonance imaging signals, there are also dynamic patterns on the brain electrical, called "micro-states", defined as global patterns of the scalp signal topology that change over time in a specific tissue fashion. That is, the brain electricity in the resting or task state can be described as several kinds of scalp voltage topological graphs (maps), which can maintain the stable state of 60-120 milliseconds, and quickly convert to other topologies, and then continue to maintain the stability. In view of its temporal resolution, the pattern of brain electrical micro-states may provide a more optimal indicator of rapid changes and therefore may better reflect the level of consciousness in patients with disturbance of consciousness, but this assumption has not been experimentally verified so far.

The Auditory oddball paradigm (audio oddball components) is widely used in electroencephalographic studies to detect residual consciousness of patients with disturbance of consciousness. In these paradigms, although the patient may be instructed to count specific target sounds heard or sounds that violate temporal rules, these paradigms rely on an assessment of the sensory response at several levels. Using an active paradigm (e.g., imagine playing tennis), partially conscious impairment patients are found to respond following instructions, which also require greater cognitive ability. Unlike previous pure tones as auditory stimuli, some studies have attempted to develop a reliable linguistic paradigm to detect neural features of semantic processing, since natural language stimuli may be more readily noticed by patients. Although neuroimaging studies have provided evidence of cortical responses to natural speech in some non-responsive patients, it has been shown that natural language stimuli activate more auditory cortical areas than disorganized stimuli, and that the results of electroencephalography are not uniform. Most electroencephalogram researches examine the N400 components of the response of neural signals to natural language narrative contents or the correlation among the testees, and find that no difference or only weak difference exists between patients with the non-response arousal syndrome and the micro-consciousness state. In the language paradigm combining stimulus-locked response and brain dynamic state, brain electricity is not adopted to assist in judging and prognosing the degree of disturbance of consciousness of a patient with disturbance of consciousness.

Disclosure of Invention

The technical problem to be solved by the present invention is to provide a system for evaluating the degree of an conscious disturbance by using brain activity related to language, a method for predicting recovery tendency of the conscious disturbance, and a storage medium.

In order to solve the above technical problem, the present invention provides a system for evaluating a degree of disturbance of consciousness, comprising: a language generator configured to generate speech material that produces auditory stimuli to a subject; an electroencephalograph configured to record an electroencephalogram signal of the subject; and the analysis device is configured to extract characteristic parameters from the electroencephalogram signals and obtain the consciousness disturbance degree of the subject according to the characteristic parameters.

In one embodiment of the invention, the stimulus comprises one or more of a word stimulus, and a sentence stimulus.

In one embodiment of the invention, the word stimulus comprises a single word that occurs at a first frequency, the word stimulus comprises a word that occurs at a second frequency, and the sentence stimulus comprises a sentence that occurs at a third frequency.

In one embodiment of the present invention, the first frequency comprises 4 hz.

In an embodiment of the invention, the second frequency comprises 2hz, the words comprise 2 words, and the 2 words comprise noun phrases.

In an embodiment of the invention, the third frequency comprises 1hz, the sentence comprises a 4-word sentence, and the 4-word sentence comprises a noun phrase and/or a motile phrase.

In an embodiment of the present invention, the characteristic parameter includes inter-trial phase consistency ITPC, which is calculated by using the following formula:

wherein f represents frequency, ITPC (f) represents trial phase consistency when the frequency of the electroencephalogram signal is f, A_k(f) The phase information of the electroencephalogram signal of the kth trial is shown, and n shows the number of total trials.

In an embodiment of the invention, the characteristic parameters include A, B, C, D micro-state parameters based on four electroencephalogram maps, and the micro-state parameters include an a state representing an occurrence probability of the a topographic map, a B state representing an occurrence probability of the B topographic map, a C state representing an occurrence probability of the C topographic map, and a D state representing an occurrence probability of the D topographic map.

In one embodiment of the invention, the micro-state parameters further comprise A-P states and L-R states, the A-P states represent the occurrence probability of front and rear topographic maps which are obtained based on A, B topographic maps in an average mode, and the L-R states represent the occurrence probability of left and right topographic maps which are obtained based on C, D topographic maps in an average mode.

In an embodiment of the present invention, the micro-state parameters further include a difference between the A-P state and the L-R state.

In an embodiment of the invention, said micro-state parameters further comprise probability weighted spatial correlation coefficient differences Δ C between said A-P states and said L-R states_oThe following formula is used for calculation:

wherein n represents the total number of recording channels of the electroencephalogram recording module, I represents the ith electroencephalogram electrode, and I_iRepresenting the voltage values measured on the kth topographic map, a-P representing front and rear topographic maps averaged based on the A, B topographic map, L-R representing averaging based on the C, D topographic mapLeft and right topographic maps, V, obtained_APRepresenting the voltage value, V, measured on said front and rear topographic maps_LRRepresenting the voltage values, p, measured on said left and right topographic maps_kThe appearance probability of the kth topographic map is shown, k represents the number of the topographic map, and the numbers k of the topographic map A, B, C, D are 1,2, 3, and 4, respectively.

In an embodiment of the present invention, the micro-state parameters further include an occurrence frequency of the A-P state, a duration of the L-R state, an inter-A-P state transition rate, and an inter-L-R state transition rate.

In an embodiment of the invention, the analysis device includes a linear discriminant analysis classifier, an input of the linear discriminant analysis classifier is the feature parameter, and an output of the linear discriminant analysis classifier is a degree of disturbance of consciousness of the subject.

In one embodiment of the invention, the degree of disturbance of consciousness in the subject includes a state of micro-consciousness and unresponsive arousal syndrome.

In an embodiment of the present invention, the method further includes: a display configured to generate a visual task that produces visual stimuli to the subject; a subject feedback button configured to receive a press by the subject, the subject pressing the subject feedback button according to the visual task.

In an embodiment of the present invention, the system further includes a prediction device, the characteristic parameter further includes inter-trial phase consistency ITPC, the input of the prediction device is one or more of the inter-trial phase consistency and the micro-status parameter, the output of the prediction device is a positive result or a negative result, the positive result indicates that the disturbance of consciousness of the subject has a tendency to recover, and the negative result indicates that the disturbance of consciousness of the subject does not have a tendency to recover.

In an embodiment of the present invention, when the stimulation speech is a word stimulation speech, the input of the prediction device includes: a difference between the A-P state and the L-R state in the micro-state parameters, and a frequency of occurrence of the A-P state; when the stimulation speech is word stimulation speech, the input of the prediction device comprises: the frequency of occurrence of the A-P states in the micro-state parameters and the inter-trial phase consistency; and when the stimulation speech is sentence stimulation speech, the input of the prediction means includes: probability-weighted spatial correlation coefficient differences between the A-P states and the L-R states in the micro-state parameters, and the frequency of occurrence of the A-P states.

In order to solve the above-mentioned problems, the present invention also provides a method for predicting a recovery tendency of an consciousness disorder, comprising: applying a stimulus to the subject; recording an electroencephalogram signal of the subject; and extracting characteristic parameters from the electroencephalogram signals, and predicting whether the consciousness disorder of the subject has a recovery tendency or not according to the characteristic parameters.

In an embodiment of the invention, the characteristic parameters include one or more of inter-trial phase consistency and micro-state parameters.

The present invention further provides a storage medium storing computer program code, which when executed by a processor performs the following steps: controlling a speech generator to generate speech material that produces auditory stimuli to a subject; controlling an electroencephalograph meter to record an electroencephalogram signal of the subject; and controlling an analysis device to extract characteristic parameters from the electroencephalogram signals, and obtaining the degree of the conscious disturbance of the subject according to the characteristic parameters and predicting whether the conscious disturbance of the subject has a recovery tendency.

The invention adopts a brand-new multi-level auditory language sequence paradigm which comprises three processing levels, namely single words, words and sentences, and compares the electroencephalogram characteristics of a passive resting state and an active task state (three language conditions). The aim of the study of the present invention is to explore the depth of speech processing in conscious impairment patients, to this extent distinguishing two different possibilities. First, the higher the state of consciousness, the deeper the depth of processing for speech stimulation may be. This hypothesis seems to be reliable, in view of the integration of words into words and sentences, persistent and integrated brain activity is often found to be associated with intentional processing. Secondly, it appears that a considerable degree of linguistic processing also occurs in unconscious conditions, as evidenced by the masking and inattentive paradigm of multiple items performed on normal subjects, including observing the brain's response to syntactic and semantic violations in unconscious conditions. Even so, it is clinically meaningful to explore the depth of unconscious processing of speech stimuli on the patient, perhaps to predict patient recovery.

This multi-level auditory language sequence paradigm enables the present invention to integrate speech tracking activity with the dynamics of brain state. The related research of the invention firstly examines whether the hierarchical structure in the processing sequence needs top-down cognitive resources and whether the hierarchical structure is carefully regulated and controlled on a healthy subject. When facing to the patient, the relevant research of the invention adopts a multivariate analysis method, combines the neural activity tracked by voice and the time dynamic characteristics of the brain macroscopic state to check the language processing depth of the patient with disturbance of consciousness, and investigates the significance of electroencephalogram measurement on consciousness judgment and prognosis. The invention trains and tests a classification algorithm which uses a data matrix derived from electroencephalogram and judgment of the degree of disturbance of consciousness in related research and is used for predicting the recovery tendency of individual patients.

Drawings

The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the principle of the invention. In the drawings:

FIG. 1 is a block diagram of a system for assessing the level of an conscious disturbance according to an embodiment of the present invention;

FIG. 2 is a schematic flow chart of a system for assessing the level of an conscious disturbance for screening patients in data analysis according to an embodiment of the present invention;

3A-3D are schematic diagrams of a multi-level linguistic paradigm and neural tracking associated with a study of the present invention;

FIGS. 4A-4B are schematic diagrams of the flow of clinical studies and auditory evoked brain activity in studies related to the present invention;

fig. 4C is the inter-trial phase consistency response results for the layered language structure individuals tested at the three task levels for the healthy control group (n-47), the micro-conscious state patient (n-42), and the unresponsive arousal syndrome (n-36) according to the study procedure shown in fig. 4A;

FIG. 4D is a trial-to-trial phase consistency response of the hierarchical language structure of an individual patient at three task levels according to the study flow shown in FIG. 4A;

FIGS. 5A-5E are schematic diagrams of global patterns of brain states obtained from studies related to the present invention;

FIG. 6 is a diagram of the brain state of healthy controls and patients under all four task conditions;

FIGS. 7A-7F are diagrams of the duration and number of occurrences of a brain state diagram obtained by an analysis device of the present invention from brain electrical signals;

FIG. 8 is the duration and number of occurrences of a brain state diagram obtained by the analysis apparatus of the present invention from brain electrical signals;

FIGS. 9A-9I are results of a correlation analysis of brain lesion volume and Δ Cp obtained in a study related to the present invention;

FIG. 10 shows the results of linear discriminant analysis between groups tested;

11A-11F are diagrams of the analysis device of the present invention obtaining a correlation of the degree of disturbance of consciousness of a subject based on a characteristic parameter;

FIGS. 12A and 12B are representations of the outcome of the prognosis and judgment using a support vector machine;

FIG. 13 is a block diagram of a system for assessing the degree of an conscious disturbance according to another embodiment of the present invention;

FIGS. 14A-14E are comparisons of classifier determinations and outcome predictions based on the brain electrical and coma recovery tables;

FIG. 15 is a comparison of the results of brain electrical versus brain electrical plus revised coma recovery scale scores and predictive performance of the results;

fig. 16A and 16B are multiple revision coma recovery scale scores across time.

Detailed Description

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings used in the description of the embodiments will be briefly introduced below. It is obvious that the drawings in the following description are only examples or embodiments of the application, from which the application can also be applied to other similar scenarios without inventive effort for a person skilled in the art. Unless otherwise apparent from the context, or otherwise indicated, like reference numbers in the figures refer to the same structure or operation.

As used in this application and the appended claims, the terms "a," "an," "the," and/or "the" are not intended to be inclusive in the singular, but rather are intended to be inclusive in the plural unless the context clearly dictates otherwise. In general, the terms "comprises" and "comprising" merely indicate that steps and elements are included which are explicitly identified, that the steps and elements do not form an exclusive list, and that a method or apparatus may include other steps or elements.

The relative arrangement of the components and steps, the numerical expressions, and numerical values set forth in these embodiments do not limit the scope of the present application unless specifically stated otherwise. Meanwhile, it should be understood that the sizes of the respective portions shown in the drawings are not drawn in an actual proportional relationship for the convenience of description. Techniques, methods, and apparatus known to those of ordinary skill in the relevant art may not be discussed in detail but are intended to be part of the specification where appropriate. In all examples shown and discussed herein, any particular value should be construed as merely illustrative, and not limiting. Thus, other examples of the exemplary embodiments may have different values. It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, further discussion thereof is not required in subsequent figures.

In the description of the present application, it is to be understood that the orientation or positional relationship indicated by the directional terms such as "front, rear, upper, lower, left, right", "lateral, vertical, horizontal" and "top, bottom", etc., are generally based on the orientation or positional relationship shown in the drawings, and are used for convenience of description and simplicity of description only, and in the case of not making a reverse description, these directional terms do not indicate and imply that the device or element being referred to must have a particular orientation or be constructed and operated in a particular orientation, and therefore, should not be considered as limiting the scope of the present application; the terms "inner and outer" refer to the inner and outer relative to the profile of the respective component itself.

Spatially relative terms, such as "above … …," "above … …," "above … …," "above," and the like, may be used herein for ease of description to describe one device or feature's spatial relationship to another device or feature as illustrated in the figures. It will be understood that the spatially relative terms are intended to encompass different orientations of the device in use or operation in addition to the orientation depicted in the figures. For example, if a device in the figures is turned over, devices described as "above" or "on" other devices or configurations would then be oriented "below" or "under" the other devices or configurations. Thus, the exemplary term "above … …" can include both an orientation of "above … …" and "below … …". The device may be otherwise variously oriented (rotated 90 degrees or at other orientations) and the spatially relative descriptors used herein interpreted accordingly.

It should be noted that the terms "first", "second", and the like are used to define the components, and are only used for convenience of distinguishing the corresponding components, and the terms have no special meanings unless otherwise stated, and therefore, the scope of protection of the present application is not to be construed as being limited. Further, although the terms used in the present application are selected from publicly known and used terms, some of the terms mentioned in the specification of the present application may be selected by the applicant at his or her discretion, the detailed meanings of which are described in relevant parts of the description herein. Further, it is required that the present application is understood not only by the actual terms used but also by the meaning of each term lying within.

The invention adopts a multi-level language processing paradigm to test the residual consciousness of the patient with disturbance of consciousness. The present invention demonstrates that two electroencephalographic-derived neural signals, speech tracking activity and global dynamic patterns, are correlated with behavioral determination of consciousness and clinical outcome. This correlation increases significantly as the language hierarchy increases. Furthermore, the multiple electroencephalographic measurements are sufficiently robust for behavioral determination and prediction of future outcomes for individual patients. Thus, this represents a new method and a new system means for clinically using electroencephalography to determine and prognose consciousness in patients with disturbance of consciousness.

During the past decade, neuroimaging and electrophysiological methods have been used to examine the state of consciousness of unresponsive patients, including functional magnetic resonance imaging, positron emission tomography, electroencephalography, and multi-channel imaging. Functional magnetic resonance imaging has many limitations, including high cost, lack of portability, and the impossibility of bedside clinical trials, while high-density electroencephalograms are more feasible to deploy at the patient bedside, facilitating longitudinal tracking of individual patients.

Previous functional magnetic resonance imaging experiments show that the brain spontaneously generates a series of dynamic and constantly changing activities and functional connections between brain regions, which are helpful for efficient information exchange between nerve cell populations; this suggests that conscious neural associations can be discovered in a dynamic process of time evolution. The present invention explores the similar discrete and dynamic states of the electroencephalogram (called "micro-states"), suggesting that the dynamic patterns of scalp potential reflect the instantaneous state of overall neural activity, possibly corresponding to changes in consciousness over time. In particular, studies related to the present invention have found that global brain activity, particularly the duration of L-R states and the number of occurrences of A-P states in language tasks, can significantly distinguish non-responsive arousal syndrome and micro-conscious state patients on a group and individual level. Furthermore, the ability to distinguish between unresponsive arousal syndrome and micro-conscious state patients by examining the dynamic activity of brain states increases with increasing task hierarchy-from rest, words to sentence conditions. The results support the notion that those electroencephalographic states are not meaningless periodic patterns, may indeed separate sensory and higher cognitive functions, and are associated with levels of consciousness and task performance (e.g., a hierarchy of language processing).

Although there have been some promising neuroimaging studies, there are few reliable active brain electrical paradigms to aid in the diagnosis and prognosis of disturbance of consciousness. The related research of the invention firstly provides an electroencephalogram evidence: the speech tracking neural responses and cortical dynamic patterns are directly related to the multilevel speech processing of the conscious disturbance patient. The present invention-related studies found that in patients with unresponsive arousal syndrome, word and sentence level responses disappeared, indicating that once unconsciousness was lost, deep speech processing failed to remain (consistent with previous studies of anesthesia and sleep). Although these responses are significantly reduced in patients with micro-conscious states, multivariate and brain state analysis indicates that language architecture continues to regulate neural processing, with some degree of deeper level processing present in patients with micro-conscious states. The increase in cortical dynamic variability in patients with micro-conscious states may explain why the word and sentence frequency response of this group is diminished in the phase-consistent spectrum. Recently, it has been argued that the reaction of word and sentence frequency reflects the influence of neural induction on the syntactic structure of mental construction, or the semantic features of a single word. However, the present invention-related research is not intended to distinguish between semantic and syntactic processing, but rather uses word and sentence frequency responses as a general standard for high-level language processing. The current result is that word rate responses remain in patients with unresponsive arousal syndrome and micro-consciousness, while responses to word rate and sentence rate decrease, further confirming that word rate and sentence rate responses reflect a deeper level of speech processing than word rate responses. In the future, the evaluation of the language understanding ability of individual patients can be facilitated by combining a plurality of electroencephalogram paradigms, including the current paradigms and syntactic and semantic violation paradigms.

The results of the related research prove the judgment potential of the brain electricity to the speech response. The decision state classification and future outcome prediction models indicate language tracking responses (e.g., trial-to-trial phase consistency at 2 and 4 Hz) and global dynamic patterns (e.g., Occurrence)_A-PAnd Δ C_ρ) Optimal decoding and prediction accuracy is required. This shows that, in combination with different analytical methods, one can mentionFor better judgment and prediction capabilities. It is noted that unlike most previous studies that predict recovery in unconscious patients, the studies related to the present invention have shown recovery predictions in patients with different states (e.g., unresponsive arousal syndrome, micro-conscious states, micro-conscious state deviation, etc.). However, it should also be noted that the studies associated with the present invention lack a detailed set of continuous behavioral measures for patients in convalescent periods. Thus, while the study of the present invention selects patients who have acquired a revised coma recovery scale more than 100 days after electroencephalographic evaluation, the study of the present invention does not suggest that electroencephalographic signals can accurately predict clinical outcome 100 days prior to performance. According to the previous study, after 6 months of onset of the conscious disturbance, 17% of the patients with non-invasive unresponsive arousal syndrome can recover consciousness, and 67% of the patients with traumatic unresponsive arousal syndrome can recover consciousness. The system evaluates whether the brain electrical model can predict behavior recovery and over what days further research is needed. However, there are several reasons why the findings of the studies related to the present invention make the studies related to the present invention confident for clinical use of electroencephalography at the bedside. First, the current paradigm has proven useful for a variety of languages, including chinese, english, and hebrew, suggesting that it may be used clinically in different linguistic contexts. Second, it is theoretically possible to use fewer electrodes and a cheaper brain electrical system to perform trial-to-trial phase consistency of the brain electrical and brain global state analysis (e.g., 16 channels) while still retaining discriminative power and clinical utility. Finally, the time frame of this paradigm is also applicable to daily bedside examination in hospitals or homes, since the duration of the experiment does not exceed 20 minutes in the case where only two language conditions (words and sentences) are involved.

FIG. 1 is a block diagram of a system for assessing the level of an conscious disturbance according to an embodiment of the present invention. Referring to fig. 1, the system 100 includes: a speech generator 110, an electroencephalograph 120, and an analysis apparatus 130. Wherein the language generator 110 is configured to generate speech material that produces auditory stimuli to the subject; the electroencephalograph 120 is configured to record an electroencephalogram signal of the subject; the analysis device 130 is configured to extract a characteristic parameter from the electroencephalogram signal, and obtain the degree of disturbance of consciousness of the subject from the characteristic parameter.

The system 100 of this embodiment and the related research process are described in detail below with reference to the drawings.

The study protocol of the invention was approved by the ethical committee of the Huashan hospital affiliated at the university of Compound denier (approval No.: HIRB-2014-281), and all the health subjects and the patient's family members signed an informed consent in advance. All subjects tested are Mandarin Chinese native. The number of samples was not determined in advance using statistical methods, but was similar to previous reports.

In this specification "subject" is a subject.

27 healthy people participated in the first electroencephalogram study, which explored Top-down attention (Top-down attentions) on the regulation of neural activity associated with voice tracking (experimental details are shown below; 15 men, average age 23.9 years, age range between 20 and 30 years). In the final analysis, 5 subjects with poor data quality were excluded, and 12 men were selected from 22 subjects with an average age of 23.73 years and an age range of 20 to 30 years.

The study of the present invention performed subject recruitment and information collection from 2016 to 2018, and the specific subject information is shown in table 1. Where GCS represents the Glasgow coma Scale and CRS-R represents the revised coma recovery Scale.

TABLE 1.2016-2018 detailed demographic and clinical information for patients recruited

Electroencephalographic data was collected from patients with Disturbance Of Consciousness (DOC) used in the study between 2016 and 2019 for 6 months. All patients were judged as either micro-conscious status (MCS) or unresponsive arousal syndrome status (UWS) based on clinical CRS-R (coma recovery) and GCS (glasgow) scales. To reduce the effect of the drug on spontaneous brain activity and wakefulness, patients were not sedated (normally midazolam) within 24 hours prior to collection.

Fig. 2 is a schematic flow chart of the system for assessing the degree of disturbance of consciousness screening a patient in data analysis according to an embodiment of the present invention.

Referring to fig. 2, electroencephalogram data of 93 patients were collected between 2016 and 2018 and 10 months, wherein some patients underwent multiple acquisitions. The data set finally contained 133 resting state brain electrical data from 89 patients and 132 task state data from 92 patients. Due to the environmental noise and the large electrical noise caused by excessive motion of the patients, a total of 62 resting data from 50 patients and 54 task state data from 43 patients were rejected. The data finally used were derived from 70 patients whose causes of disturbance of consciousness included stroke (36, 51.43%), traumatic brain injury (31, 44.29%) and ischemic-hypoxic encephalopathy (3, 4.29%) (see table 1 in particular). There were 71 available resting state brain data from these patients, 54 consciousness disorder patients (48 men, mean age 49.3 years, age range between 17 and 75 years), 78 task state brain data from 60 consciousness disorder patients (52 men, mean age 47.8 years, age range between 9 and 68 years). The number of recordings was slightly different for each patient.

Both the attention and clinical studies of the present invention were double blind (both subjects and experimenters were blind). Neither the subjects involved in the study nor the experimenter collecting the data were aware of the true purpose of the study. Both the interviewee and interviewer receiving follow-up assessments were blinded to the behavioral and electroencephalographic judgment of the hospital. In addition, the experimenter collecting the electroencephalogram data does not participate in data analysis.

In particular, in clinical studies, behavioural measurements and electroencephalography and magnetic resonance imaging data analysis were performed by two groups of experimenters, respectively, without any knowledge at all. Two researchers analyzed and modeled electroencephalography and magnetic resonance imaging data, but were blinded to the patient's results measurements. Four other investigators performed clinical assessments of patient behavior, but did not involve electroencephalography and magnetic resonance imaging data analysis.

Patients were classified into two groups, micro-conscious state and non-responsive arousal syndrome, based on behavioral assessment.

Patients with unresponsive arousal syndrome are awake but without any conscious behavioral manifestations. This type of patient may have their eyes open, have a simple reflex, and may switch between awake and sleep states. Patients with a micro-conscious state retain some residual consciousness and exhibit faint but reproducible signs of consciousness. In addition, the patient with micro-consciousness state can be further divided into two subsets according to the complexity of behaviors. Micro-conscious state + the patient may make a higher level of behavioral response, exhibiting compliance, vocalization, or nonfunctional communication. Micro-conscious states-the level of behavioral response of the patient is lower and can be identified by eye-tracking, localization of noxious stimuli, accidental reactions associated with environmental stimuli, and the like. Finally, a patient who is able to functionally communicate and/or manipulate different objects is determined to be out of micro-consciousness (EMCS). Clinical classification and behavioral scoring are performed by experienced physicians on the day of detection, usually prior to electroencephalography.

In some embodiments, the stimuli generated by language generator 110 shown in FIG. 1 include one or more of word stimuli, and sentence stimuli.

In some embodiments, the word stimulus comprises a single word that occurs at a first frequency, the word stimulus comprises a word that occurs at a second frequency, and the sentence stimulus comprises a sentence that occurs at a third frequency.

In some embodiments, the first frequency comprises 4 hertz.

In some embodiments, the second frequency comprises 2 hertz, the words comprise 2-word words, and the 2-word words comprise noun phrases.

In some embodiments, the third frequency comprises 1 hertz, the sentences comprise 4-word sentences, and the 4-word sentences comprise noun phrases and/or motile phrases.

The speech material used in the study of the present invention is an auditory chinese material, which originates from previous work and is slightly altered. Chinese auditory stimuli include 1 (words only), 2 (words and phrases), or 3 (words, phrases and sentences) language hierarchies. The speech material is synthesized online by a free text-to-speech generating engine (http:// ai. basic. com/tech/speech/tts).

Table 2 shows chinese material contained in the stimuli generated by the speech generator in the system of one embodiment of the present invention.

TABLE 2 Chinese materials

The present invention first defines 50 4-word sentences composed of noun phrases and motile phrases, as shown by "4-word sentences" in table 2. 200 characters forming a sentence pass through a character-voice generating engine one by one to obtain Chinese pronunciation. Subsequently, based on these phrases, 20 64 word sequences were generated, each word of the sequence being randomly selected from a 200 word library, and the intervals between the utterances were manually adjusted so that each word took 250 milliseconds, and the entire sequence was 16 seconds long.

From the 50 4-word sentences in the condition of the terms in table 2, 50 nouns are extracted to form a word bank of 2-word words, as shown by "2 words" in table 2. 32 words are randomly selected from the word stock each time to form a 16-second word sequence, and 20 words are formed. The pronunciations of the words are synthesized word by the engine on line to avoid continuous reading.

Similar to word stimulation, 16 sentence sequences consisting of 16 seconds were randomly chosen at a time from a library of 50 sentences. This randomization process was repeated 20 times.

3A-3D are schematic diagrams of the multi-level linguistic paradigm and neural tracking associated with the present invention. Fig. 3A is an exemplary diagram of the stimulation generated by the language generator in the system for assessing the degree of disturbance of consciousness according to an embodiment of the present invention. Referring to FIG. 3A, the sentence condition contains 3 semantic levels, including words that occur at a frequency of 4Hz, words that occur at a frequency of 2Hz, and sentences that occur at a frequency of 1 Hz. The word condition includes only words and word frequencies, while the word condition has only word frequencies. In this embodiment, the first frequency is 4 hertz, the second frequency is 2 hertz, and the third frequency is 1 hertz.

In each recording, 30 choices (each sequence having an occurrence frequency of 2 or less) are put back from the corresponding 20 presynthesized sequences under each condition and are joined into continuous speech having a total length of 480 seconds.

This test was done in an electro-acoustic shielded room and The experimental stimuli were presented by a psychtools kit based on MATLAB software (R2015b, The MathWorks inc., USA).

Note that the experiment was designed with a full factorial design, containing two factors: note (two levels, note or ignore) and language condition (two levels, single word or sentence stimulus). Thus, there are a total of four blocks with different task conditions, respectively:

note that the word: attention is paid to word audio, and visual tasks which are simultaneously performed are ignored;

note that sentence: note the audio of the sentence, ignore the visual task that is going on at the same time;

ignore word: focusing on the visual task, and neglecting the audio of the character which appears simultaneously;

ignore sentence: note that visual, simultaneous sentence audio is ignored.

FIG. 3B is a schematic illustration of an attention test of a healthy participant in a system for assessing the level of an disturbance of consciousness in accordance with an embodiment of the present invention. The subject was asked to attend or ignore a visual attention task in a different block while presenting 8 minutes of chinese speech material, as in the auditory flow of fig. 3B. The auditory stream was adapted from previous work, consisting of single syllabic words of chinese, containing one (word) or three (word, word and sentence) language levels, as shown in fig. 3A.

The auditory flow started 20 seconds after the first visual trial for each experimental block began and ended before the end of the last visual trial and was played through two speakers about 80 cm away from the subject's ears at the display, at a sound pressure level of about 65 db. In general, the auditory stream under each condition consists of 30 sets of 16 second long chinese sequences with no significant separation between the sequences. The audio for each tile is played for 8 minutes continuously.

In this embodiment, the system for assessing the degree of disturbance of consciousness according to the present invention further includes: a display and a subject feedback button. Wherein the display is configured to generate a visual task that produces a visual stimulus to the subject; the subject feedback button is configured to receive a press by the subject, the subject pressing the subject feedback button according to the visual task. Referring to fig. 3B, in the visual task, the display is used to display corresponding graphics, symbols, letters or numbers, and the mouse button is used as the subject feedback button.

The visual stimulus was presented on a 23 inch liquid crystal display approximately 60 cm from the subject. A cross-point of fixation is presented within 1.5 seconds of the start of the visual test, followed by a presentation consisting of a graphic, a symbol or symbols, and numbers. After 4 seconds, a pattern matrix is given. The pattern matrix consists of 5 shapes (isosceles right triangle, equilateral triangle, square, pentagon, hexagon), with a random total number (24 ± 2) and 4 colors (blue, green, yellow, magenta). In the case of visual attention, the subject is required to react to the matrix within 12 seconds, by pressing the left or right arrow key to indicate whether the previous expression was correct (the number of a particular shape in the matrix is greater or less than a given number). The assignment of keys (with or without the prior statement) was balanced between different trials. Immediately after the response, visual feedback was given for 2.5 seconds to indicate whether the response was correct. The next trial was started after 3-6.1 seconds intervals. Each experimental block contained 32 trials for about 10 minutes.

Under visual neglect conditions, the shape matrix is always present for 7 seconds in each trial, as the subject does not need to react. The subject is asked to focus on the audio and ignore the visual test. After completing a block, the subject is asked to determine whether the words/sentences in the test list have been played.

The order of the four task blocks was randomized and balanced across the trials. The electroencephalogram data is continuously acquired and divided into 16-second time periods. To obtain clean data, the present invention-related study excluded trials involving noise, extreme motion, and blinking. The average number of trials for the attention word, attention sentence, negligence word and negligence sentence conditions were 28.7, 28.2, 28.8 and 27.4, respectively.

It should be noted that in the behavioral analysis of the visual task, the behavioral differences between the attention sentence and the attention word condition were not significant (accuracy: 75.71 ± 1.79% vs. 76.56 ± 2.52%, p ═ 0.796; reaction time: 6.99 ± 0.24 vs. 6.84 ± 0.33 sec; p ═ 0.518; paired t-test).

Fig. 4A-4B are schematic diagrams of the flow of clinical studies and auditory evoked brain activity in studies related to the present invention. Wherein fig. 4A is an exemplary flow of an consciousness-impaired patient study. Referring to fig. 4A, according to this exemplary procedure, in a study on the same day, enrolled patients with impaired consciousness were first behaviorally scored using the revised coma recovery scale (CRS-R) and Glasgow Coma Scale (GCS), with electroencephalographic recordings beginning at a resting state of 5 minutes, followed by short breaks and randomized three verbal auditory chunks between subjects.

This test was performed in a hospital or The like, and experimental stimulation was presented by a psychtools kit based on MATLAB software (R2015b, The MathWorks inc., USA).

First, an electroencephalogram was measured at rest for 5 minutes at the beginning of each recording. After a 2 minute rest period, three groups of stimuli were presented. These groups correspond to three different semantic levels of 8-minute chinese audio sequences: word, phrase, and sentence conditions. Before each task group, a short introduction was played to guide the subject to quiet and listen carefully, which was also synthesized using the same online text-to-speech engine. To reduce ambient noise, researchers have delivered sound stimuli through headphones at a volume of about 65 db, and the participants have worn a pair of sound-insulating earmuffs over the headphones.

The order of the task conditions is random and balanced, controlled by a random function in MATLAB. In addition, the stimulation sequence under each task condition was also adjusted between the subjects.

Note that the electroencephalogram data for the study was acquired by a 64-channel electroencephalogram system (actiCHamp, Brain Products GmbH, Germany), and the electroencephalogram data for the clinical study was acquired by a 257-channel electroencephalogram system (GES 300 or GES 400, Electrical Geodesics Inc., USA), with a sample rate of 1000 Hz. During electroencephalogram data acquisition, FCz (attention research) or Cz (clinical research) is referenced on line, and the electrode impedance is kept below 5 kilo-ohm (attention research) or 20 kilo-ohm (clinical research).

The 64-channel and/or 257-channel electroencephalography system is one embodiment of the electroencephalograph 120 shown in fig. 1.

In view of the noisy electrical noise of the recording environment, the more involuntary extra movements of the patient, the study applied the following criteria on each data channel, examining the proportion of bad duration during each recording (257 data channels in the inter-trial phase consistency analysis; 204 electrodes in the brain state analysis, since the electrodes placed on the cheek and neck were excluded from the analysis first):

1) gradient: the instantaneous change in voltage should not exceed 30 microvolts/millisecond;

2) extreme amplitude: within a sliding window of 200 milliseconds, the difference between the maximum and minimum voltage values should not exceed 120 microvolts, the window sliding in 10 millisecond steps;

3) absolute amplitude: the absolute value of the amplitude value should not exceed 100 microvolts;

4) low signal: the difference between the maximum and minimum voltage values should not be less than 1 microvolt within a sliding window of 100 milliseconds, with the window sliding in 10 millisecond steps.

Data points that exceed the quality criteria are marked as bad data points and a period of 200 milliseconds before and after that point is marked as a bad period. Electrodes containing more than 20% of the total recording time bad period will be marked as bad electrodes, and the current recording data with more than 70 bad channels will be discarded.

In some embodiments, the feature parameters extracted from the brain electrical signal by the analysis device 130 shown in fig. 1 include inter-trial phase consistency. In these embodiments, a single trial brain wave is first runAnd performing discrete Fourier transform on the data without adding a smooth window. The result for the kth trial (k ═ 1,2, …, n) is labeled X_k(f) Then the phase information should be A_k(f)＝∠X_k(f) In that respect Then the Inter-Trial Phase Coherence (ITPC, Inter-Trial Phase Coherence) is defined by the following equation:

In the inter-trial phase consistency analysis, electroencephalogram data was preprocessed using a Brain vision Analyzer (2.0.1, Brain Products, GmbH, Germany) with the following analysis steps: band-pass filtering at 0.1-40 Hz in combination with 50 Hz notch; then, checking a bad electrode and a time period, and carrying out interpolation recovery on the bad data point; the data were re-referenced as an average of all electrodes; using independent component analysis to remove blink and saccade artifacts; the continuous data was divided into 16 second trial segments and the sampling rate was reduced to 50 hz.

In the inter-trial phase consistency analysis, the related research of the invention carries out statistical test on individual trial and group trial. A one-sided precision test was used at the individual level. The trial phase consistency value between 0.2 and 5 Hz is divided into 77 sections according to 1/16 Hz step, and the original distribution is formed by the trial phase consistency value under the frequency irrelevant to the Chinese structure. The statistical significance of the response at the target frequency (precision P) is thus the probability that the target frequency response differs from a zero distribution (non-target frequency; number of non-target frequencies tested in three cases: 76 times for words, 75 times for words, 74 times). At the population level, the target frequency was compared to the mean of 4 frequency bins above and below the target frequency, 0.25 hz. Statistical significance is the difference between the response of the target frequency and the response of its neighboring frequencies (one-sided paired t-test). For the classification results, the significance of decoding performance was examined using a one-sided paired t-test, with a random level of 0.5.

In some embodiments, the characteristic parameters extracted from the brain electrical signal by the analysis device 130 shown in FIG. 1 include micro-state parameters.

In brain electrical microstate analysis, brain electrical data was preprocessed using MATLAB-based eglab toolbox (version 14.1.1), the analysis steps are as follows: electrodes for removing the neck and face; performing band-pass filtering (0.2 to 40 Hz) on the reserved 204 electrodes, checking bad electrodes and time periods, and performing interpolation recovery on bad data points; using independent component analysis to remove blink and saccade artifacts; dividing continuous data into test time sections of 2 seconds, and removing bad test times; the data is re-referenced and additionally subjected to 2-20 hz bandpass filtering.

Brain status was analyzed using the MicrositeAnalysis kit based on MATLAB and EEGLAB (version 0.3, software downloaded freely on http:// www.thomaskoenig.ch/index. php/software/microstates-in-EEGLAB /). Under each condition, the electroencephalogram micro-state is obtained based on k-means clustering analysis of a electroencephalogram topographic map when the activity variation of the whole brain electroencephalogram electrode is maximum. Based on cross-validation and global interpretation variance, the number of topograms for the micro-states is limited to 4. The study selects 4 electroencephalography maps (a to D) of healthy persons under each condition as templates for all the groups tested to obtain the best global interpretation variance and stability.

In some embodiments, the feature parameters extracted from the brain electrical signals by the analysis device 130 include A, B, C, D-based micro-state parameters of four brain electrical maps, including an a-state representing the occurrence probability of an a-topographic map, a B-state representing the occurrence probability of a B-topographic map, a C-state representing the occurrence probability of a C-topographic map, and a D-state representing the occurrence probability of a D-topographic map.

The study of the present invention calculated the micro-State Occurrence Probability (Brain State Probability), Mean Duration (Mean Duration), Mean number of occurrences (Mean occupancy) and Mean Transition Probability (Mean Transition Probability) of all the groups tested under all the test conditions based on the template. To summarize the 4 brain states for a single subject, the correlation study of the present invention calculated the probability-weighted spatial correlation coefficient differenceIs Δ C_ρ。

In some embodiments, the feature parameters extracted from the brain electrical signal by the analysis device 130 further include A-P states and L-R states, the A-P states represent the probability of occurrence of the front and back topographic maps averaged based on the A, B topographic maps, and the L-R states represent the probability of occurrence of the left and right topographic maps averaged based on the C, D topographic maps. In these embodiments, the class 4 topographical templates are further summarized as anterior-posterior (A-P) and left-right (L-R) topographical maps, averaged based on the A, B and C, D topographical maps, respectively. That is, front and rear maps are averaged based on the A, B map, and left and right maps are averaged based on the C, D map.

In some embodiments, the micro-regime parameters also include the difference between the A-P regime and the L-R regime, denoted as Δ Prohealth.

For each test, the spatial correlation of each topographic map is obtained by calculating the spatial pearson correlation coefficient of the topographic map to be tested and the template topographic map (front-back and left-right topographic maps). The difference in correlation coefficient (ac) with both templates on each patient indicates the degree of similarity to healthy subjects.

Each difference in spatial correlation corresponds to a spatial pearson correlation, which is calculated as follows:

where Δ C is the difference between the correlation coefficients of the two templates, n is the number of electrodes, I is the voltage value measured on each topographic map, and V_APAnd V_LRMeasured voltage values on the front-back and left-right templates, respectively, i being the ith electrode. The front-back template is a front-back topographic map, and the left-right template is a left-right topographic map.

Correspondingly, the probability weighted spatial correlation coefficient difference Δ C between the A-P state and the L-R state is calculated according to the following formula_ρ：

Where ρ is the probability of occurrence of a given topographic map (topographic maps A, B, C and D), ρ_kThe probability of occurrence of the kth topographic map is shown, k is the kth topographic map, and the numbers k of the topographic maps A, B, C, D are 1,2, 3, and 4, respectively.

In some embodiments, the micro-state parameters also include the frequency of Occurrence of the A-P state Occurrence_A-PDuration of L-R state_L-RTransition rate Transition between A-P states_A-PAnd L-R inter-state Transition rate Transition_L-R。

In brain state analysis, the main effect among groups for each condition was examined using multi-factor analysis of variance for probabilities of 4 topographic maps. Differences between different conditions within each group were examined using repeated measures analysis of variance, groups (3 levels: healthy controls, micro-consciousness status, unresponsive arousal syndrome) were interclass variables, and electroencephalogram characteristic parameters and tasks (3 levels: words, sentences) were intraclass variables. Pairwise comparisons between groups were tested by Bonferroni corrected one-way anova. In analyzing follow-up patients, the first and last differences in brain electrical activity of patients were analyzed using the friedman test. Statistical significance of the classification and prediction analysis results was tested using the chi-square test (Fisher exact test). All data distributions were assumed to be normal, but were not formally tested.

The present invention correlation study uses independently synthesized isochronous, 4Hz Chinese word sequences to build a multi-level linguistic structure, as shown in FIG. 3A. Auditory sequences include three language levels, namely monosyllabic words, two-word words, and four-word sentences. Monosyllabic words are presented at a constant frequency (250 milliseconds per single word), which means that corresponding neural tracking of words, phrases and sentences can be tracked at different frequencies.

The invention firstly tests whether the top-down attention is necessary for the multi-level language structure electroencephalogram response. Studies related to the present invention 22 healthy human subjects were recruited to complete an attention task. The subject either notices the auditory stimulus (notice condition), i.e., a list of words or a sequence of sentences, or performs a visual task while the auditory stimulus is presented and ignored, as shown in fig. 3B.

Fig. 3C is the mean results of phase consistency between trials of healthy participants under four conditions of attention experiment. Wherein the top two panels represent the one-sided paired t-test under attentive conditions: t is t_{Word-4 Hz}(21)＝10.11,P_{Character (Chinese character)}-_4Hz＝8×10^-10；t_Sentence-_1Hz(21)＝6.11,P_Sentence-_1Hz＝2.3×10^-6；t_Sentence-_2Hz(21)＝7.26,P_{Sentence-2 Hz}＝1.9×10^-7；t_{Sentence-4 Hz}(21)＝11.1,P_{Sentence-4 Hz}＝1.4×10^-10. The following two graphs represent the one-sided paired t-test under neglected conditions: t is t_{Word-4 Hz}(21)＝8.22,P_{Word-4 Hz}＝2.67×10^-8；t_Sentence-_1Hz(21)＝4,P_Sentence-_1Hz＝3.2×10^-4；t_Sentence-_2Hz(21)＝4.14,P_Sentence-_2Hz＝2.3×10^-4；t_Sentence-_4Hz(21)＝8.58,P_Sentence-_4Hz＝1.3×10-⁸。

Fig. 3D is a graph of healthy participants paying attention to and ignoring comparison results under four conditions of attention to the experiment. Referring to fig. 3D, based on the two-sample two-tailed t-test: t is t_{Word-4 Hz}(21)＝0.42,P_{Word-4 Hz}＝0.678；t_Sentence-_4Hz(21)＝1.58,P_Sentence-_4Hz＝0.128,t_Sentence-_2Hz(21)＝5.41,P_Sentence-_2Hz＝2.3×10^-5,t_Sentence-_1Hz(21)＝2.66,P_{Sentence-1 Hz}0.015. Where light dots represent individuals and dark dots represent averages. Error bars represent s.e.m. full version: n.s., P>0.05；*,P<0.05；***,P<0.001。

Under note and ignore conditions, the present invention correlation study found that there was a significant 4Hz response in the inter-trial phase consistency (ITPC) spectrum under both word list and sentence conditions, as shown in FIG. 3C, as compared to baseline. Note, however, that after moving to visual stimuli, the inter-trial phase consistency values at 1 and 2hz were significantly diminished under the sentence condition, as shown in fig. 3D, note that the paired t test for neglect resulted in,P_1Hz0.015 and P_2Hz＝2.3×10^-5. Thus, these results indicate that there is automation of processing on the basis of word tracking (4 hz), and partial attentional regulation in neural processing of higher-level linguistic structures, i.e., words and sentences (2 and 1hz, respectively).

Based on these results, the present invention-related studies assume that residual consciousness of consciousness-impaired patients can be reflected by the strength of the speech-tracking response, particularly for neural tracking of high-level linguistic structures, i.e., words and sentences (inter-trial phase consistency measures). To test this hypothesis, the study related to the present invention examined brain responses to sentence sequences in 42 patients with micro-consciousness, 36 patients with unresponsive arousal syndrome, and 47 healthy controls, with details of the patients shown in table 1 and a screening of the patients shown in fig. 2.

As shown in fig. 4A, after the determination of the disturbance of consciousness using the revised coma-recovering scale (CRS-R) and Glasgow Coma Scale (GCS) (classification of the micro-consciousness state and the unresponsive arousal syndrome after the determination of the revised coma-recovering scale), the resting-state electroencephalogram was recorded for 5 minutes first at the start of each recording. After 2 minutes of rest, 3 8-minute chinese stimulation blocks containing different language levels (word list, word sequence, and sentence sequence) are presented.

FIG. 4B is the result of phase consistency between the group mean trials under word, sentence conditions according to the study flow shown in FIG. 4A. Data were obtained from healthy controls (n: 47), micro-consciousness (n: 42), and non-responsive arousal syndrome (n: 36), each plotted from top to bottom. Single-sided paired t-test. The results are as follows:

word conditions are as follows: t is t_{4 Hz-healthy group}(46)＝8.24,P_{4 Hz-healthy group}＝1.3×10^-10；t_4Hz-MCS(41)＝5.52,P_4Hz-MCS＝2.1×10^-6；t_4Hz-UWS(35)＝3.78,P_4Hz-UWS＝5.8×10^-4。

Word conditions: t is t_{4 Hz-healthy group}(46)＝8.78,P_{4 Hz-healthy group}＝2.1×10^-11；t_4Hz-MCS(41)＝5.36,P_4Hz-MCS＝3.5×10^-6；t_4Hz-UWS(35)＝3.87,P_4Hz-UWS＝4.5×10^-4；t_{2 Hz-healthy group}(46)＝7.25,P_{2 Hz-healthy group}＝3.8×10^-9；t_2Hz-MCS(41)＝1.70,P_2Hz-MCS＝0.097,t_2Hz-UWS(35)＝0.15,P_2Hz-UWS＝0.881。

Sentence conditions are as follows: t is t_{4 Hz-healthy group}(46)＝8.35,P_{4 Hz-healthy group}＝9.1×10^-11；t_4Hz-MCS(41)＝5.40,P_4Hz-MCS＝2.2×10^-6；t_4Hz-UWS(35)＝3.93,P_4Hz-UWS＝3.8×10^-4；t_{2 Hz-healthy group}(46)＝6.62,P_{2 Hz-healthy group}＝3.4×10^-8；t_2Hz-MCS(41)＝1.88,P_2Hz-MCS＝0.068,t_2Hz-UWS(35)＝0.39,P_2Hz-UWS＝0.702；t_{1 Hz-healthy group}(46)＝4.48,P_{1 Hz-healthy group}＝4.9×10^-5,t_1Hz-MCS(41)＝0.58,P_1Hz-MCS＝0.567,t_1Hz-UWS(35)＝-0.61,P_1Hz-UWS＝0.546。

Fig. 4C is the inter-trial phase consistency response results for the layered language structure individuals tested at the three task levels for the healthy control group (n-47), the patients with micro-consciousness (n-42), and the unresponsive arousal syndrome (n-36) according to the study procedure shown in fig. 4A. In each inset, the left-hand point represents the trial-to-trial phase consistency values of individuals at the target frequency (1, 2, 4 hz), and the right-hand point represents their neighboring individual mean values. Solid black dots represent the overall average. n.s., P > 0.1; p < 0.1; p < 0.001; single-sided paired t-test: the exact statistical values are described above with respect to FIG. 4B.

The above results show that the phase consistency intensity between electroencephalograms increases gradually from unresponsive arousal syndrome to micro-conscious states, matching the level of increased behavioral responsiveness quantified by the revised coma-recovery scale, compared to healthy controls. Specifically, as shown in the 3 panels in the left column of FIG. 4B and FIG. 4C, word-level tracking, as measured by 4Hz inter-trial phase consistency, was significant in the healthy control group, the micro-conscious state group, and the unresponsive arousal syndrome group (P)_4Hz-health＝1.3×10^-10；P_4Hz-MCS＝2.1×10^-6；P_4Hz-UWS＝5.8×10^-4(ii) a Paired sample t test). As shown in the 3 panels in the middle column of FIG. 4B and FIG. 4C, word level tracking of the 2Hz trial-to-trial phase consistency measure was significant in the healthy control group, marginal in the micro-conscious state group, and not significant in the unresponsive arousal syndrome group (P)_{2 Hz-healthy}＝3.8×10^-9；P_2Hz-MCS＝0.097；P_2Hz-UWS0.881; paired sample t test). As shown in the 3 panels in the right column of FIG. 4B and FIG. 4C, the sentence level tracking for phase consistency between 1Hz trials was significant in the healthy control group and not significant in the micro-conscious and unresponsive arousal syndrome groups (P)_{1 Hz-healthy}＝4.9×10^-5；P_1Hz-MCS＝0.567；P_1Hz-UWS0.546; paired sample t test).

Fig. 4D is a trial-to-trial phase consistency response of the hierarchical language structure of individual patients at three task levels according to the study procedure shown in fig. 4A. Where each bar represents a reaction from one subject. The dots represent significance (exact P < 0.05; one-sided exact test, statistical significance of inter-trial phase-consistency responses at target frequency (exact P) is the probability that the target frequency response differs from a zero distribution, which consists of responses at all non-target frequencies).

Notably, while the phase consistency between trials at 1 or 2hz was not significantly different between the micro-conscious state and the unresponsive arousal syndrome groups as shown in figure 4B, there were some significant differences at the individual level as shown in figure 4D. That is, 11 patients with a micro-conscious state and 4 patients with unresponsive arousal syndrome showed significant phase consistency between trials at 1 or 2 hertz, which may indicate that these patients are still conscious. In fact, 6 of these 15 patients (5 micro-consciousness state and 1 non-reactive arousal syndrome; 40%) had a significant improvement in judgment of the degree of disturbance of consciousness 100 days after electroencephalogram recording (prediction of the results can also be seen in the classification results).

The brain is spontaneously and regularly active at rest and during cognitive activities, and this dynamic pattern is considered a neural feature of consciousness. Thus, the present invention-related studies evaluated the second hypothesis that residual consciousness can be characterized by monitoring the dynamic patterns of brain states, as these brain dynamics are associated with different cognitive states. To this end, the study related to the present invention quantifies the spatial and temporal dynamics of brain activity in healthy control groups and patients by examining the characteristics (e.g., probability of occurrence, number of occurrences, duration, and rate of transitions between states) of the global pattern of scalp potential maps (also referred to as "microstations") under conditions of increased levels of four languages.

Fig. 5A-5E are schematic diagrams of global patterns of brain states obtained from studies related to the present invention. Where fig. 5A shows that the average accuracy and average global interpretation variance (GEV) of the cross-validation (CV) criteria varies with the number of brain state graphs, indicating that the optimal number of inter-trial cluster graphs is 4.

Fig. 5B is a group mean brain state plot for healthy control groups under task conditions (averaged over all three tasks, n-47 subjects), sorted by brain state probability. The shade of the color represents the relative potential distribution.

Fig. 5C is a probability distribution of three sets of four topographical maps for four cases. Multivariate analysis of variance: at rest, F (6,200) ═ 5.176, P ═ 5.7 × 10^-5(ii) a Word, F (6,240) ═ 13.543, P ═ 3.1 × 10^-13(ii) a The term F (6,240) 14.258, P6.9 × 10^-14(ii) a In sentence, F (6,240) is 12.56, and P is 2.6 × 10^-12. The lines represent the mean and the shaded areas represent the s.e.m.

FIG. 5D is the Δ Cp (probability weighted spatial correlation difference) of the A-P and L-R topographs for each group and condition. One-way anova, Bonferroni correction: at rest, P_{Health group-MCS}＝3.7×10^-11,P_{Health group-UWS}＝2.4×10^-9(ii) a Word, P_{Health group-MCS}＝2.9×10^-19,P_{Health group-UWS}＝3.8×10^-22(ii) a Word, P_{Health group-MCS}＝2.1×10^-15,P_{Health group-UWS}＝9.4×10^-21,P_MCS-UWS0.037; sentence, P_{Health group-MCS}＝1.7×10^-15,P_{Health group-UWS}＝9.9×10^-22,P_MCS-UWS0.014. The light dots represent individuals. Black dots represent the average. Error bars represent s.e.m.

Fig. 5E is a comparison of Δ C ρ for micro-consciousness state and unresponsive arousal syndrome. This difference matches well with the language hierarchy from the resting state to the sentence condition. The broken line indicates the statistical significance between the patient groups in each case. One-way anova, Bonferroni correction: p_{At rest}＝1,P_{Character (Chinese character)}＝0.33,P_Word＝0.037,P_Sentence＝0.014.Panels c-e:n_{Health group-rest}＝34,n_{Health group-task}＝47,n_MCS-rest＝41,n_MCS-task＝42,n_UWS-rest＝30,n _UWS-task36. Full edition: n.s., P>0.05；*,P<0.05；***,P<0.001。

As shown in fig. 5A, population level clustering identified the best cluster for four different populations and conditions, achieved the highest cross-validation criteria, accounting for approximately 80% of the variation.

As shown in fig. 5B, the spatial configuration of the four state plots of the healthy control group is highly consistent with the four state plots described in the previous studies. Then, the relevant study of the present invention labeled and ranked four groups of graphs according to the probability of brain state appearance in resting state for healthy control group. Specifically, as shown in fig. 5B, state diagram a is the central maximum of the forehead, state diagram B is the symmetrical forehead rest orientation, state diagram C is the left forehead rest orientation, and state diagram D is the right forehead rest orientation.

Fig. 6 is a diagram of the brain state of healthy controls and patients under all four task conditions. Number of subjects: n is_Health-rest＝34,n_Health-task＝47,n_MCS-rest＝41,n_MCS-task＝42,n_UWS-rest＝30,n _UWS-task36. Putting on the fence: anterior-posterior (A-P) and left-right (L-R) template plots obtained from healthy controls. Three columns at the bottom: the original four plots for each group under each condition (fig. A, B, C, D).

Studies using both electroencephalography and functional magnetic resonance imaging recordings have shown that brain states a and B are more closely related to attention and highlight networks because their corresponding Blood Oxygen Level Dependent (BOLD) activations are located in the anterior cingulate cortex and parietal region, and that C and D states are related to the auditory and visual sensory networks because their corresponding blood oxygen level dependent signals are located in the bilateral temporal and exorphic visual regions. The present invention studies predict that the higher the level of consciousness, the greater the likelihood of activation of the high level cognitive neural network, i.e., graphs A and B (shown in FIG. 5B, the anterior-posterior graph, defined as the A-P state). Meanwhile, the studies related to the present invention predict that hypoconsciousness will involve sensory areas at a lower level, corresponding to panels C and D (left-right panels are defined as L-R states as shown in fig. 5B). As shown in FIG. 5C, multivariate analysis of variance (MANOVA) showed that for these four cases, healthy controls showed a pattern of high probability for the A-P state and low probability for the L-R state. The patient groups showed the opposite pattern, with a low probability of the A-P state and a high probability of the L-R state.

The related research of the invention further tests the difference of the dynamic characteristics of the brain electrical micro-state between the micro-consciousness state and the non-reactive arousal syndrome patients. For each patient, the correlation study of the present invention calculated the probability weighted spatial correlation coefficient difference Δ C ρ between the A-P and L-R states. This difference reflects the spatial similarity of the map of the patient group and the template map of the healthy control group and serves as an indicator of residual consciousness.

As shown in fig. 5D and 5E, the present inventors found that the difference in Δ C ρ between the group of micro-consciousness states and the group of unresponsive arousal syndrome gradually increased from rest to the word, sentence condition as the language hierarchy increased. Among them, as shown in fig. 5E, at the word and sentence level (excluding rest and word level), the group of micro-consciousness states was significantly higher than the group of non-reactive arousal syndrome (P)_{Word and phrase}＝0.037；P_Sentence0.014; one-way analysis of variance, Bonferroni check). This indicates that the probability of the A-P state (frontal lobe network) increases and the probability of the L-R state (sensory network) decreases. In addition, the difference in Δ C ρ between the micro-consciousness state and the unresponsive arousal syndrome under the word and period conditions is significantly greater than that under the rest condition and the word condition (word pair, t76 ═ 2.29, P ═ 0.03; sentence pair, t76 ═ 3.14, P ═ 0.002; double sample two-tailed t test). Thus, positive and higherThe ac ρ may characterize residual consciousness.

Further, the present inventors studied whether the difference in probability of a state is due to Duration (time for which the state remains stable) or due to frequency of occurrence of each state Occurence (how many times the state occurs in one second).

Fig. 7A to 7F are the duration and the number of occurrences of a brain state diagram obtained by the analysis device of the present invention from brain electrical signals. Wherein the duration is shown in fig. 7A. Wherein, left edition (L-R): duration of L-R status (n) of healthy control group_{At rest}＝34,n_Task47), patients with micro-consciousness (n)_{At rest}＝41,n_Task42), patients with unresponsive arousal syndrome (n)_{At rest}＝30,n_Task36). One-way anova, Bonferroni correction: at rest, P_{Health group-MCS}＝5.6×10^-10,P_{Health group-UWS}＝5.8×10^-10,P _MCS-UWS1 is ═ 1; word, P_{Health group-MCS}＝2.7×10^-8,P_{Health group-UWS}＝8.2×10^-13,P_MCS-UWS0.083; word, P_{Health group-MCS}＝1.1×10^-8,P_{Health group-UWS}＝1.05×10^-14,P_MCS-UWS0.016; sentence, P_{Health group-MCS}＝1.1×10^-9,P_{Health group-UWS}＝1.1×10^-15,P_MCS-UWS0.017. The boxes represent IQR, the center points represent intermediate values, and the boxes represent 1.5 × IQR. The dots represent outliers. Right edition (follow-up): restoring the duration of the initial (circular) and final (real) electroencephalographic L-R states of a patient (restoration: n)_Task＝7,n_Task8) and duration of unrecovered patients (unrecovered: n is_Task＝5,n_Task7). The line (solid line: + ve; dashed line: -ve) represents the initial and final electroencephalographic recording of the individual. Friedman test: rest, x² _Recovery＝3.57,P_Recovery＝0.059,χ² _{Is not recovered}＝1.8,P_{Is not recovered}0.18; word, chi² _Recovery＝8.0,P_Recovery＝0.005,χ² _{Is not recovered}＝0.14,P_{Is not recovered}＝0.705(ii) a Word, chi² _Recovery＝8.0,P_Recovery＝0.005,χ² _{Is not recovered}＝3.57,P_{Is not recovered}0.059; sentence, x² _Recovery＝8.0,P_Recovery＝0.005,χ² _{Is not recovered}＝0.14,P_{Is not recovered}＝0.705。

FIG. 7B is a comparison of L-R state duration for the micro-conscious state group and the unresponsive arousal syndrome group under four conditions. The increase in the difference in the duration of the L-R state between the group of micro-conscious states and the group of unresponsive arousal syndrome paralleled the language hierarchy. The broken line represents the statistical significance of the difference between the L-R state duration groups under each condition.

Figure 7C is a graph comparing the difference in duration of initial and final recorded L-R status between patients recovering and patients not recovering under four conditions.

FIGS. 7D-7F illustrate the occurrence of the A-P state, which is in the same format as FIGS. 7A-7C. FIGS. 7D and 7E show the A-P difference, one-way ANOVA, Bonferroni correction: at rest, P_{Health group-MCS}＝1.1×10^-12,P_{Health group-UWS}＝1.2×10^-11,P _MCS-UWS1 is ═ 1; word, P_{Health group-MCS}＝6.1×10^-15,P_{Health group-UWS}＝7.1×10^-19,P_MCS-UWS0.156; word, P_{Health group-MCS}＝3.4×10^-15,P_{Health group-UWS}＝8.4×10^-21,P_MCS-UWS0.028; sentence, P_{Health group-MCS}＝3.6×10^-16,P_{Health group-UWS}＝5.4×10^-21,P_MCS-UWS0.063. Fig. 7D and 7F show the follow-up differences, and fig. 7F is a summary of the four sub-graph follow-up columns in fig. 7D, friedman test: rest, x² _Recovery＝3.57,P_Recovery＝0.059,χ² _{Is not recovered}＝0.2,P_{Is not recovered}0.655; word, chi² _Recovery＝8.0,P_Recovery＝0.005,χ² _{Is not recovered}＝1.29,P_{Is not recovered}0.257; word, chi² _Recovery＝8.0,P_Recovery＝0.005,χ² _{Is not recovered}＝1.29,P_{Is not recovered}0.257; sentence, x² _Recovery＝8.0,P_Recovery＝0.005,χ² _{Is not recovered}＝0.14,P_{Is not recovered}0.705. Full version of n.s., P>0.05；*,P<0.05；**,P<0.01；***,P<0.001。

According to the results of the study of the present invention, as shown in FIGS. 7A and 7B, the difference in probability between the micro consciousness state group and the unresponsive arousal syndrome group under the word and sentence conditions can be attributed to the fact that the L-R state duration is short for the unresponsive arousal syndrome group in the micro consciousness state group, which is considered to reflect the sensory network (P-R state)_{Word group}＝0.016；P_Sentence0.017 percent; one-way anova, Bonferroni's check), as shown in fig. 7D and 7E, the frequency of occurrence of a-P state is high, which is related to the frontal lobe network (P)_{Word group}＝0.028；P_Sentence0.063; one-way analysis of variance, Bonferroni check). Importantly, as shown in fig. 7B and 7E, the increase in significant difference between the micro-conscious state group and the unresponsive arousal syndrome group matched the increase in conditional language level.

Fig. 8 is the duration and the number of occurrences of a brain state diagram obtained by the analysis device of the present invention from brain electrical signals. Wherein the first row a is a healthy control group (n)_{At rest}＝34,n_Task47), group of micro-consciousness states (n)_{At rest}＝41,n_Task42) and the unresponsive arousal syndrome group (n)_{At rest}＝30,n_Task36) a-P state duration under four task conditions. Note that there is no difference between the three groups. The boxes represent IQR, the center points represent intermediate values, and the boxes represent 1.5 × IQR. The dots represent outliers.

The second row b is the number of occurrences of L-R status under four task conditions in the healthy control group, the group of micro-conscious states, and the group of unresponsive arousal syndrome. There were no differences between the three groups.

The third row c is a healthy control group (light gray outline circle; n)_{At rest}＝34,n_Task47), the patient was recovered (+ ve, light dots; n is_{At rest}＝11,n_Task19), patients with unrecoverable micro-consciousness state (-ve MCS, dark gray outline circle; n is_{At rest}＝30,n_Task23) and unrecovered nothingPatients with reactive arousal syndrome (-ve UWS, dark circle dot; n_{At rest}＝30,n_Task36) duration under four conditions. One-way anova, Bonferroni correction: healthy vs. + ve, P_{At rest}＝0.001,P_{Character (Chinese character)}＝0.018,P_Word＝0.013,P_Sentence＝0.008；+ve vs.-ve MCS,P_{Character (Chinese character)}＝0.012,P_Word＝0.009,P_Sentence＝0.002。

The fourth row d indicates the occurrence of the A-P state in healthy controls, convalescent patients, patients without recovery of micro-consciousness state, and patients without recovery of non-reactive arousal syndrome. One-way anova, Bonferroni correction: healthy vs. + ve, P_{At rest}＝6.8×10^-5,P_{Character (Chinese character)}＝8.1×10^-7,P_Word＝5.7×10^-7,P_Sentence＝1.8×10^-7；+ve vs.-ve MCS,P_{Character (Chinese character)}＝0.046,P_{P word}＝0.048,P_Sentence＝0.026。

In the third and fourth rows, the light dots represent individuals. Black circles represent the average. Error bars represent s.e.m.

Throughout the full version of fig. 8: p < 0.05; p < 0.01; p < 0.001.

As shown in FIG. 8, there is no significant difference between the duration of the A-P state and the occurrence time of the L-R state.

If the duration and/or occurrence of the plot does reflect the intensity of the patient's residual consciousness, the study associated with the present invention should observe the corresponding changes before and after the patient recovers. In the subset of subjects with multiple electroencephalography recordings (12 out of 54 patients had multiple resting state recordings, 15 out of 60 patients had multiple language task recordings), with recovery, as shown in fig. 7A and 7C, the duration of the L-R state became shorter, and as shown in fig. 7D and 7F, the appearance of the a-P state increased. As shown in fig. 7A, 7C, 7D, 7F, this change was not observed in the non-recovered patients. The distribution of the duration of the A-P state and the number of occurrences of the L-R state in healthy control groups, convalescent and non-convalescent micro-conscious states, and non-reactive arousal syndrome patients is shown in FIG. 8.

To exclude the possibility that the difference in state space patterns between the two groups of patients resulted from brain injury differences, the present invention-related study analyzed 27 brain injury patients (17 micro-conscious states and 10 unresponsive arousal syndromes) obtained from the database who had received a structural magnetic resonance imaging scan on the day of electroencephalography.

On the day of electroencephalography, Magnetic Resonance (MRI) structural images of 27 patients (16 males, mean age 44.6 years; 9 to 68 years) were collected. MRI data were acquired using either a 3T (21 cases, Siemens magnetic Verio, Germany, using turbo spin-echo sequence) or 1.5T (6 cases, GE Signal EXCITE twisted Zoom, USA, using a fast spin-echo sequence) magnetic resonance scanner. After indexes such as structural image contrast, head movement and damage detection reliability are checked, brain damage sites are determined on T2W1 images of 21 patients, FLAIR images of 4 patients and T1W1 images of 2 patients in related researches, damage areas are extracted under the cooperation of doctors, and the volume is calculated by multiplying the sum of all voxels in the areas by the size of the voxels. The correlation study of the present invention further calculated the pearson correlation between ac ρ and lesion volume. Magnetic resonance imaging data were analyzed using MATLAB (2017b, MathWorks, USA) and ITK-SNAP (version number: 3.8).

FIGS. 9A-9I are results of a correlation analysis of brain lesion volume and Δ Cp obtained in a study related to the present invention. Fig. 9A is a comparison of brain lesion volume in patients with micro-consciousness (n 17) and without arousal syndrome (n 10). Black dots and error bars represent mean and s.e.m, respectively. t is t₂₅0.64, 0.53, two-sample two-tailed t-test.

Fig. 9B is a correlation of brain lesion volume to ac ρ under three task conditions. Pearson correlation test (two tails), n_MCS＝17，n_UWS＝10.

Fig. 9C is a patient example (patient 7): nuclear magnetic resonance data and figures for stroke patients without brain damage.

Fig. 9D is a patient example (patient 17): nuclear magnetic resonance data and figures for patients with extensive traumatic brain injury.

Fig. 9E is a comparison of delta Proavailability for stroke patients without brain injury and patients with traumatic brain injury, as shown in fig. 9C and 9D. P1: patient 7, P2: patient 17.

FIG. 9F is Δ C ρ; the format is the same as fig. 9E.

Figure 9G is nuclear magnetic resonance data and graphs for stroke patients with brain injury (example patient, patient 2). The left box represents the first electroencephalogram recording in the unrecovered state. The box on the right represents the last electroencephalogram recording in the recovery state.

Fig. 9H is a comparison of delta Prohealth of patient 2's electroencephalography recording in the first, unrecovered state and the last, restored state.

Fig. 9I is Δ C ρ, the same format as fig. 9H. W: word, P: word, S: and (4) sentence(s). F: initial record, L: and finally recording. The percentage under each graph represents the probability of each graph appearing.

The results show that there was no significant difference in lesion volume between patients with micro-consciousness status and unresponsive arousal syndrome as shown in fig. 9A, and no significant correlation between brain lesion volume and ac ρ at three task levels as shown in fig. 9B. For example, as shown in fig. 9C, the value of Δ C ρ is low in the patient 07 without brain damage, as shown in fig. 9D, while the value of Δ C ρ is relatively high in the patient 17 with large brain damage. Furthermore, if the studies associated with the present invention assume that the size of the brain injury does not change significantly during the recovery period, the observed spatial map changes before and after the recovery period, as shown in FIGS. 7A-7F, also help to exclude the possibility that the brain state measured using the spatial map reflects purely a potential brain injury. The spatial map change of the patient (ID:2) is shown in FIGS. 9G-9I.

In some embodiments, the analysis device 130 shown in fig. 1 includes a linear discriminant analysis classifier, the input of which is the feature parameter, and the output of which is the degree of the conscious disturbance of the subject.

In some embodiments, the degree of disturbance of consciousness in the subject includes a state of micro-consciousness and unresponsive arousal syndrome.

The tested groups are distinguished in a binary mode. Since there were 3 groups of subjects (healthy control group, patients with micro-consciousness state, patients with unresponsive arousal syndrome), pairwise linear discriminant analysis classification training was performed at each target frequency under each task, and decoding was achieved as follows: 1) inputting 257 trial phase consistency values which are characterized by each channel of the electroencephalogram; 2) in each comparison (comparing the phase consistency values between two groups of trials with certain frequency under certain task conditions), 4/5 trials are randomly selected to be classified as a training set, and the rest 1/5 is a test set; 3) 5-fold cross-validation of the training set, that is, for each fold, the subjects eligible for 4/5 were classified and validated on the training set of 1/5; 4) finally, based on the probability classification of the independent test set, taking the sum of the Area Under the operating characteristic Curve (AUC) of the receiver as the performance of the classifier; 5) steps 2 to 4 were repeated 100 times to obtain the area under the operating characteristic curve for the average reception of the two groups at each frequency under each condition.

Next, the present invention-related study applied multivariate pattern analysis to 1,2, and 4hz trial-to-trial phase consistency for all available electrodes to classify patient groups (linear discriminant analysis).

FIG. 10 shows the results of linear discriminant analysis between the groups tested. It shows the decoding performance of multivariate mode analysis using phase-consistency linear discriminant analysis at 1,2, and 4 hertz. Single tail t-test, compared to 0.5: micro-consciousness State on non-reactive arousal syndrome, t_{Word-4 Hz}(99)＝-4.31,P_{Word-4 Hz}＝0.99998；t_{Word-4 Hz}(99)＝2.75,P_{Word-4 Hz}＝3.6×10^-3,t_{Word-2 Hz}(99)＝-2.75,P_{Word-2 Hz}＝0.996,t_{Sentence-4 Hz}(99)＝6.52,P_{Sentence-4 Hz}＝1.5×10^-9,t_{Sentence-2 Hz}(99)＝4.8,P_{Sentence-2 Hz}＝2.4×10^-6,t_{Sentence-1 Hz}(99)＝2.02,P_{Sentence-1 Hz}0.029; state of health versus micro-consciousness, t_{Word-4 Hz}＝17.63,P_{Word-4 Hz}＝1.3×10^-32,t_{Word-4 Hz}(99)＝17.45,P_{Word-4 Hz}＝2.8×10^-32,t_{Word-2 Hz}(99)＝30.04,P_{Word-2 Hz}＝7.6×10-⁵²,t_{Sentence-4 Hz}(99)＝17.25,P_{Sentence-4 Hz}＝6.7×10^-32,t_{Sentence-2 Hz}(99)＝38.61,P_{Sentence-2 Hz}＝8.7×10-⁶²,t_{Sentence-1 Hz}(99)＝6.31,P_sentence-1Hz＝4×10-⁹(ii) a Healthy versus unresponsive arousal syndrome, t_{Word-4 Hz}(99)＝23.75,P_{Word-4 Hz}＝5.6×10^-43,t_{Word-4 Hz}(99)＝25.79,P_{Word-4 Hz}＝5×10^-46,t_{Word-2 Hz}(99)＝31.06,P_{Word-2 Hz}＝3.8×10^-53,t_{Sentence-4 Hz}(99)＝31.25,P_{Sentence-4 Hz}＝2.2×10^-53,t_{Sentence-2 Hz}(99)＝34.71,P_{Sentence-2 Hz}＝1.6×10^-57,t_{Sentence-1 Hz}(99)＝17.36,P_{Sentence-1 Hz}＝4.1×10^-32. The boxes represent the quartile range (IQR), the central bar represents the median, and the box must represent 1.5 × IQR. The "+" symbol indicates an abnormal value. The horizontal dotted line indicates a random level. Full edition: n.s., P>0.1；～,P<0.1；*,P<0.05。

Fig. 10 shows successful decoding of a patient group. Even the micro-conscious state group and the unresponsive arousal syndrome group are clearly distinguished, especially when listening to a sentence sequence.

The present study first identified the state of consciousness of the subject using a taxonomic analysis. The data set used for classification analysis was subjected to the following data culling criteria: 1) the disease course of the patient with disturbance of consciousness is shorter than 3 months; 2) patients received Deep electrical Stimulation (DBS) treatment within the previous 120 days; 3) the patient may have an unstable state of consciousness due to other unexpected diseases (e.g., lung infection). The final data set used contained 47 healthy subjects, 31 subjects with micro-consciousness and 30 subjects with unresponsive arousal syndrome. These feature combinations were used to train a three-class linear discriminant analysis classifier to distinguish healthy controls, patients with micro-consciousness status, and patients with unresponsive arousal syndrome. When using the electroencephalogram indices of all three levels of linguistic tasks, there are 893 possible feature combinations in total (period conditions contain 3 inter-trial phase consistency indices and 6 brain state indices, yielding 2⁹-1 ═ 511 feature combinations; word and word conditions yield 255 and 127 feature combinations, respectively). For each classification, to avoid model overfitting, only one of the 893 feature combinations was selected for submission to the model. All steps were cross-validated (leave one out). Regularization classificationThe classifier first searches for the optimal combination of features in each task and then calculates the classification probability for each individual being tested. To avoid model overfitting, only selected combinations of features are input in the final linear discriminant analysis. The covariance matrix is estimated using a regularized linear discriminant analysis. Cross-validation relies on a leave-one-out approach with 108 permutations. In view of the biased effect of unequal class sizes in linear discriminant analysis classification, the correlation study of the present invention does not rely on non-uniform prior probabilities for class sizes, but assumes that all classes have the same number of samples. To match the number of patients with unresponsive arousal syndrome, 29 healthy persons and 29 patients with micro-conscious states were randomly selected for classification, and this replacement procedure was repeated 2000 times. For each feature combination model fit, the accuracy of the classification was averaged over 2000 permutations. The average accuracy of the classification allows the correlation studies of the present invention to determine the optimal feature combinations. The single subject is tested under the optimal combination of eigenvalues, and the maximum probability obtained by 2000 permutations is taken as the final result of the classification.

The multiple measurements of brain activity by the present invention correlation study diverted the present correlation study from population-level analysis to attempts at personalized judgment and prediction. The related research of the invention firstly makes a linear discriminant analysis classifier of three classifications, and in order to leave a cross validation, the linear discriminant analysis classifier is used for judging whether an individual is tested, and the individual is tested as a healthy person, a micro-consciousness state with disturbance of consciousness for at least 3 months and a patient with unresponsive arousal syndrome. The inputs to the classifier are a plurality of electroencephalogram measurements including inter-trial phase consistency (3 features: inter-trial phase consistency values at 1,2 and 4 hertz) and global dynamic patterns of brain activity (6 features: difference Δ Probasic of A-P and L-R states, Probability weighted spatial correlation coefficient difference Δ Cp between A-P and L-R states, frequency of Occurrence of A-P states Occurence_A-PDuration of L-R state_L-RTransition rate Transition between A-P states_A-PAnd L-R inter-state Transition rate Transition_L-R)。

Fig. 11A to 11F are diagrams related to the degree of disturbance of consciousness of a subject obtained by the analysis device of the present invention based on the characteristic parameters. Fig. 11A shows a program flow for classification and prediction using a plurality of electroencephalogram characteristics. As shown in fig. 11A, the regularized classifier is used to search for the optimal feature combination in each task first, and calculate the classification probability for each test. To avoid model overfitting, only selected combinations of features are input in the final linear discriminant analysis. All steps were cross-validated (leave one out).

FIG. 11B shows a confusion matrix for decision awareness classification generated by linear discriminant analysis. The classifier is trained according to electroencephalogram indexes extracted from sentence tasks. The pie chart shows the mismatch between the judgment of the level of disturbance of consciousness and the results of the patient.

FIG. 11C shows the use of the revised coma recovery Table totals for outcome prediction. Checking a chi square: AUC 70%, χ²7.2, P0.016 and 74% accuracy.

FIG. 11D shows the prognostic validity of the model as assessed using the revised coma restitution scale total score. The classifier was cross-validated on a data set of 38 patients and then tested for generalization capability on a new data set of 25 patients. Checking a chi square: AUC of 31%, χ²4.6, P0.049, and 28% accuracy. Among these 38 patients, data sets 2016.7-2018.10 in FIG. 2 were subjected to a series of screenings to obtain 38 patients, 15 of which were positive and 23 of which were negative. Also shown in FIG. 11A, the 38 patients were the sum of 15 + ve and 23-ve patients who were "predictive".

With respect to the new data set comprising the 25 patients, the study employed another independent data set collected between month 10 2018 and month 6 2019, comprising 25 collections, of which there were 12 states of micro-consciousness and 13 unresponsive arousal syndromes; 13 males, with an average age of 39.9 years, ranged between 18 and 69 years. Details of the patients enrolled in this independent dataset are shown in table 3.

Table 3: 2018-novel Collection of patients enrolled in 2019 with detailed demographic and clinical information

Clinical studies 61 healthy subjects were also enrolled in the local community for experimental control (20 males, mean age 31.3 years, age range between 22 and 65 years). There were 34 cases of resting state data that passed the data quality test (8 men, mean age 25.8 years, age range between 22 and 50 years) and 47 cases of task state data (17 men, mean age 31.1 years, age range between 22 and 58 years).

FIG. 11E shows the performance of the prediction of results using electroencephalogram (left edition, chi-square test: AUC 77%, chi%²＝11.5,P＝9.2×10^-4 Accuracy 76%) and comparison of individual patient predicted and actual outcomes (right panel). Points above the threshold (grey line, prediction score ═ 0.1), diamonds, represent patients who predicted positive results (+ ve), others represent patients who predicted negative results (-ve).

FIG. 11F shows the prognostic validity of the brain-based model. The classifier was trained on the original data set and generalized to a new data set including the 25 patient data described above. Checking a chi square: AUC 80%, χ²When P is 8.8, P is 0.005 and accuracy is 80%. Fig. 11F is the same format as fig. 11E. g: group (d); ω: and (4) normalizing the coefficients.

The data referred to in FIGS. 11A, 11D, 11F, and the right columns of FIGS. 14A through 14E are new data sets.

FIG. 11B shows the confusion matrix generated by linear discriminant analysis. By Δ Cp, Transition_A-P、ITPC_1Hz、ITPC_2HzThe sentence condition classification for the input electroencephalogram features works best. Using Chi-Square test to estimate the performance of the classifier, it was shown that the classification effect was significant (χ)²＝95.84,P＝7.6×10^-20 Accuracy 75%). The classification accuracy of the decoder for healthy control group, patients with micro-consciousness state and patients with non-reactive arousal syndrome is 89%, 58% and 70%, respectively, which are far higher than the random level of 33%.

FIG. 1 shows a schematic view of a2A and 12B are the judgment and prognosis outcome presentation using a support vector machine. Fig. 12A shows a judgment-aware classification confusion matrix generated by the cross-validation support vector machine. The combination of features used by the sentence task is [ Δ Cp + Duration [ ]_L-R+Occurrence_A-P+ITPC_1Hz+ITPC_2Hz+ITPC_4Hz]。

FIG. 12B shows the performance of the support vector machine classifier using the best feature combination for outcome prediction on training data. Left: accuracy of result prediction using brain waves for 38 brain electrical recordings (15 patients positive in result). And (3) right: comparison of individual prediction and true outcome. The left side of the dotted line is the unresponsive arousal syndrome patient and the right side is the micro-conscious state patient. The points above the threshold (gray line, prediction score 0.3) represent predicted positive results, while others represent predicted negative results. Patients with negative actual results are marked with light colored dots, patients with positive actual results are marked with light colored diamonds, and patients with complete arousal are marked with solid line diamonds. The combination of features used: word conditions are as follows: [ Delta Probasic + Duration [ ]_L-R+Transition_A-P]The term: [ Delta Probasic + Occurence_A-P+Duration_L-R+Transition_L-R+ITPC_4Hz]And sentence condition: [ Occurrence_A-P+Duration_L-R+Transition_L-R+ITPC_1Hz+ITPC_2Hz]。

Referring to fig. 12A, a Support Vector Machine (SVM) verified the high decoding accuracy of the algorithm, with 96%, 65% and 73% accuracy for healthy control, patients with micro-consciousness state and patients with unresponsive arousal syndrome, respectively.

Although a proportion of patients with unresponsive arousal syndrome are classified as micro-conscious (30%, 9/30), it is likely that these patients have some degree of consciousness and cannot be detected using the revised coma-recovery scale in the performance assessment. Notably, a greater proportion of patients with unresponsive arousal syndrome who may be misdiagnosed (33.3%, 3/9) had positive outcomes (either fully awakened or exhibited behavioral elevations after electroencephalography) than were consented to be judged to be unresponsive arousal syndrome by both the revised coma recovery scale classifier and the electroencephalography classifier (9.5%, 2/21). On the contrary, the classifier also judged some patients with micro-consciousness state as unresponsive arousal syndrome (12 out of 31 cases). A lower percentage of patients with a possibly misdiagnosed micro-conscious state (25%, 12/3) had positive outcomes than patients (44.4%, 18/8) judged to be in a micro-conscious state by revision of the coma recovery scale and electroencephalography measurements. However, it should be noted in the study related to the present invention that the number of patients in these groups may be too small to allow statistical analysis of the differences between groups in these proportions.

Fig. 13 is a block diagram of a system for assessing the level of an conscious disturbance according to another embodiment of the present invention. Referring to fig. 10, the system 101 of this embodiment differs from the system 100 shown in fig. 1 in that the system 101 includes a prediction device 140 in addition to a language generator 110, an electroencephalograph 120, and an analysis device 130.

In some embodiments, the characteristic parameters further include inter-trial phase consistency ITPC, the input to the prediction means 140 is one or more of inter-trial phase consistency and the micro-status parameter, and the output of the prediction means is a positive result indicating that the disturbance of consciousness of the subject has a tendency to recover or a negative result indicating that the disturbance of consciousness of the subject does not have a tendency to recover.

In some embodiments, when the stimulation speech is a word stimulation speech, the input of the prediction device 140 includes: the difference between the A-P and L-R states in the micro-state parameters, Δ Probasic, and the frequency of Occurrence of the A-P states, Occurence_A-P(ii) a When the stimulus speech is a word stimulus speech, the input of the prediction means 140 includes: frequency of Occurrence Occurrence of A-P states in micro-state parameters_A-PAnd inter-trial phase consistency ITPC; when the stimulated speech is sentence-stimulated speech, the input of the prediction means 140 includes: probability weighted spatial correlation coefficient difference Δ Cp between A-P and L-R states in micro-state parameters, and frequency of Occurrence of A-P states Occurrence_A-P。

The system 101 of this embodiment and the related research process are described in detail below with reference to the drawings.

In order to predict the prognosis of patients, patients with behavior follow-up visit 100 days after electroencephalogram recording are selected for related research. The recovery tendency of these patients is predicted using the prediction device 140 of the present invention. Each patient was marked as a positive or negative outcome. The judgment of the level of consciousness disorder of the patient can be further classified into four categories of UWS/VS, MCS-, MCS + and EMCS according to the level of consciousness. A positive result is defined as the judgment of the degree of disturbance of consciousness at the time of follow-up visit progressing, and a negative result is maintained or returned. Predictive analysis used data from 38 patients, 15 of whom were positive (10 MCS, 5 UWS), who were fully aroused or showed significantly improved behavioral signs in subsequent measurements, and 23 negative (7 MCS, 16 UWS). Likewise, the process of consciousness state classification is also used to classify the results of the individuals. The three classifications (healthy control group, positive outcome patient, and negative outcome patient) were performed, and the corresponding normalization coefficients (ω) were 1 (healthy control group), 0.5 (positive outcome), and 0 (negative outcome). The study normalized the classification probability (P) obtained under the optimal feature value combination condition, and then the weighted average thereof was taken as the prediction score (Φ ═ Σ P · ω) under the test condition. Since the prediction scores fluctuate to some extent under different task conditions of a single tested object, the prediction scores under the conditions of 3 tasks are averaged finally.

The present invention also verifies the generalization capability of the predictive device 140. External validation (generalization ability) was performed using another independent sample (25 patients) containing 15 positive-result patients (5 MCS, 10 UWS) and 10 negative-result patients (7 MCS, 3 UWS) as shown in fig. 11D, 11F and fig. 14A-14E. First, cross-validation training was performed on a data set of 38 patients, and then a linear discriminant analysis classifier for electroencephalogram indices for outcome prediction was tested on a new data set of 25 patients. Using the behavior characteristics: the revised coma-recovery scale total score and 6 subscales were subjected to similar results prediction and summary analysis procedures (1, auditory [0-4 ]; 2, visual [0-5 ]; 3, exercise [0-6 ]; 4, verbal exercise [0-3 ]; 5, communication [0-2 ]; 6, arousal [0-3 ]). For the revised coma recovery scale total score classifier, the present correlation study also calculated opportunistic performance by repeating the same generalization 100 times using the shuffled result tags of the test dataset.

FIGS. 14A-14E are comparisons of classifier determinations and outcome predictions based on the brain electrical and coma recovery tables. Classification models (linear discriminant analysis) were cross-validated on the first 38 patients (left) and then tested on a new dataset of 25 patients (right) (no retraining).

Fig. 14A shows a result of judgment of the degree of disturbance of consciousness using the revised total score of the coma-restoration chart. Left: performance of the revised coma recovery scale score (total score) on the results prediction. And (3) right: and (4) carrying out generalization test on the model.

Fig. 14B shows a performance of the consciousness deterioration degree judgment using the revised visual subscale of the coma-restoring scale. Left: the revised coma restitution scale scores the performance of the best features (visual subscales) on the result predictions. And (3) right: and (4) carrying out generalization test on the model.

FIG. 14C shows the model prediction performance using electroencephalography under word conditions (7 features: 1 trial-to-trial phase consistency and 6 micro-states). Left: the optimal feature combination is Δ Probasic + Occurence_A-P. And (3) right: and (4) carrying out generalization test on the model.

Fig. 14D shows the expression of the judgment of the degree of disturbance of consciousness using the revised coma-recovering table, and the same number of features as those of the electroencephalogram model are used here. Left: the optimal combination of features is a vision and arousal subscale. And (3) right: and (4) carrying out generalization test on the model.

FIG. 14E shows a comparison of the results predictive performance using all 7 features of the brain and the revised coma recovery scale under word conditions using standard linear discriminant analysis without feature selection (total score and 6 scores: auditory, visual, motor, verbal motor, communication and arousal). Generalization of electroencephalography under two other task conditions (the term: AUC 89%, χ)²13.1, 5.5 × 10-4, 84% accuracy; AUC 93%, chi²13.1, 5.5 × 10-4, 84% accuracy; chi-square test) is similar to the results under the word condition.

As shown in fig. 14E, direct comparisons of the prediction of the brain electrical and revised coma recovery scale scores and generalization thereof using linear discriminant analysis were also examined without finding the optimal combination of features.

The input features of the training classifier are the values of the electroencephalogram or revised coma restitution scale (total score and 6-subscale) indices. The label for each test (sample) is either positive or negative. Under all task conditions, the related studies of the present invention used 5-fold cross validation, randomly drawn samples, and sorted by label.

The related research of the invention draws an ROC curve of the prediction score, and the area under the curve is measured for estimating the single task and the task mean value prediction result. The optimal threshold for the prediction is determined by the point on the ROC curve where the sum of sensitivity and specificity is maximal. The corresponding prediction threshold after normalization is 0.1, as shown with reference to fig. 11E and 11F. Patients with a prediction score above the threshold are determined to be a positive prognosis. And comparing the prediction label of the patient with the actual result of follow-up judgment, and calculating the prediction accuracy.

The studies associated with the present invention also examined whether electroencephalographic recordings could predict the subsequent recovery of consciousness in individual patients. 38 patients with multiple measurements were included in this analysis. The total score was based on the revised coma recovery scale 6 months after the conscious disturbance was diagnosed, with a positive prognosis for 15 patients, indicating + ve (see figure 2 for patient selection). The present invention correlation study used the same cross-validation method as described above to construct a linear discriminant analysis classifier, this time with the goal of isolating 15 patients with positive results and 23 patients with negative results. Referring to FIG. 11C, the results show that the total score of the revised coma-recovery Scale can be partially predicted (AUC 70%, χ)²7.2, P0.016, chi fang test). Referring to fig. 11E, the prediction using the electroencephalogram measurement value is better (AUC 77%, χ)²＝11.5，P＝9.2×10^-4Checking by a chi-square method; FIG. 11E to the left; sensitivity: brain electrical contrast revised coma recovery scale, P0.07, McNemar test). The best electroencephalogram predictive ability was obtained by averaging the expression of the word, word and sentence conditions with a sensitivity of 87% and a specificity of 70%, to correctly predict 13 of 15 positive patients and 16 of 23 negative patients.

The invention also comprises a method for predicting the recovery tendency of the disturbance of consciousness, which comprises the following steps:

step S1: applying a stimulus to the subject;

step S2: recording the electroencephalogram signal of a subject; and

step S3: extracting characteristic parameters from the electroencephalogram signals, and predicting whether the consciousness disorder of the subject has a recovery tendency or not according to the characteristic parameters.

In some embodiments, the feature parameters extracted in step S3 include one or more of inter-trial phase consistency and micro-state parameters.

Step S1 may be performed using the language generator 110 shown in fig. 13; performing step S2 with the electroencephalograph 120; step S3 is performed using the analysis means 130 and the prediction means 140. The relevant description can be used to describe the prediction method.

The best feature combination predicted at each task level is: delta Proavailability + Occurence under word conditions_A-POccurrence under term conditions_A-P+ITPC_4HzΔ Cp + Occurence under sentence condition_A-PAs shown in table 4.

Table 4: and (4) comparing the prediction of the scoring result of the electroencephalogram and electroencephalogram combined revision coma recovery scale under different task conditions. n is_+ve＝15,n _-ve23; and (6) checking the chi-square.

The relevant study of the present invention then examined the prognostic power of individual patients by calculating the predictive scores of the patients for unresponsive arousal syndrome and micro-conscious states based on the clinical outcome of the revised coma recovery scale. The prediction results showed high prediction accuracy for both groups (81% unresponsive arousal syndrome and 71% micro-consciousness state; individual prediction scores are shown on the right of fig. 11E). Thus, whether a single revised coma recovery scale measurement is used in a study to which the present invention relates, or multiple revised coma recovery scale measurements are used in another study, the use of electroencephalography measurements is better than behavioral observations taken alone.

It is noteworthy that although the number of subjects in the data set used to predict the model is relatively small (38 cases), the optimal feature combination for classification is only 2 or 3 features, indicating that the model is less likely to be overfitting. However, to ensure the reliability of the results, the related studies of the present invention cross-validated the same dataset using different classifier support vector machines. This additional analysis confirmed the results of the studies related to the present invention, showing that the electroencephalogram has significant predictive accuracy, as shown in fig. 12B (χ)²＝15.4，P＝1.6×10^-4Accuracy ═ 82%, chi-square test).

Most importantly, to further test the external validity and generalization ability of the model of the present invention-related study, they were tested (without retraining) on a new data set comprising 25 additional patients (12 micro-conscious states and 13 non-reactive wake syndrome states, table 3). The classifier trained using the previous dataset of the same feature combination (linear discriminant analysis) showed higher prediction accuracy in both the positive and negative groups of results for both sample sets, as shown in FIG. 11F (left, χ)²8.8, P0.005, 80% accuracy, chi fang check; individual prediction scores in the right panel). However, the classifier trained using the total score of the revised coma recovery Scale, which is similarly generalized, has much lower prediction accuracy, as shown in FIG. 11D (χ @)²4.6, P0.049, accuracy 28%, chi fang test). As a comparison, when the relevant studies of the present invention disturb the result labels of the test dataset, the classification accuracy is 50%; the decision classifier (i.e., the micro-consciousness state versus unresponsive arousal syndrome state) was successfully generalized to a new data set after a revised coma recovery scale scoring training for the first 38 patients, as shown in fig. 14A (χ)²＝21.3，P＝2.7×10^-6The accuracy rate is 96%, chi fang test).

Is the electroencephalogram measurement more predictive accuracy due to the use of more electroencephalogram features? To assess this, the present invention-related studies extended behavioral measures including total scores and six sub-scores (e.g., auditory, visual, motor, verbal, communication, and arousal) and used the same model to find the best revised coma-recovery scale feature combination for prognosis. After cross-validation of the first 38 patients, the best prediction was obtained using only the visual subscale, as shown on the left of fig. 14B, but the classifier still failed to generalize to the new 25 patient data set, as shown on the right of fig. 14B.

As shown in fig. 14C and 14D, a direct comparison of the revised coma recovery tables and the predicted performance of the brain electrical recordings under the word conditions (both with 7 features) gave similar results. In addition, the related studies of the present invention used standard linear discriminant analysis models prior to classification, without feature selection, to test the outcome prediction. As shown in fig. 14E, the results indicate that the use of electroencephalogram indicators has better generalization capability than the use of the revised coma recovery scale features.

Finally, as shown in FIG. 11C, the confusion matrix generated by the revised coma recovery scale score shows that a significant proportion of patients with positive results (53%, 8/15; 3 states of micro-consciousness and 5 non-responsive arousal syndromes) were mispredicted as negative results. As shown in fig. 11E, the electroencephalography task-based evaluation of the relevant study of the present invention may help in a more accurate determination than the revised coma recovery scale because the classifier constructed using electroencephalography measurements of the relevant study of the present invention is able to significantly predict the future second classification result of the revised coma recovery scale: of the 8 patients, 6 were classified as positive by the electroencephalogram classifier. On the basis, the relevant research of the invention combines the revised version coma recovery table and the characteristics of the electroencephalogram and submits the combined result to the model. The results show that at the word and word level, the classification accuracy using both electroencephalogram and the revised coma recovery table is slightly higher than the classification accuracy using only electroencephalogram signals.

FIG. 15 is a comparison of the results of brain electrical versus brain electrical plus revised coma recovery scale scores and the predicted performance. Wherein, the uplink is: a confusion matrix for the outcome prediction of the brain electrical score. The following actions: a confusion matrix combining the results predictions of the brain electrical and revised coma recovery scale scores. The classification accuracy at each level is shown in FIG. 15 and Table 4.

In the classification, the best combination of features does include the revised coma recovery scale, which may indicate that the predictive model may use both the electroencephalogram and the revised coma recovery scale. In conclusion, these results confirm the prognostic power of electroencephalographic indicators recorded during speech hearing and suggest that these electroencephalographic indicators can supplement clinical demographics and behavioral decisions.

The outcome of the patient in the prediction is based on a single revised coma recovery scale, which may lead to misdiagnosis due to changes in the level of consciousness over time (e.g., morning and afternoon during the day and on different days).

Fig. 16A and 16B are multiple revision coma recovery scale scores across time. Fig. 16A shows an individual patient (n-15). In each inset, the dots represent the word revised coma recovery scale score and the gray lines represent the generalized linear model fit of all scores over the entire time period. Day 0 and the vertical dashed line represent the day of the first electroencephalogram recording. Fig. 16B shows a comparison of the revised coma recovery scale score for the day of the week (2.67 days on average) and the day of electroencephalography recording. The light line indicates the score of the individual patient. The black line represents the average. There was no significant difference between the two scores (n-15, t)₁₄0.899, 0.384; two sample two tail t-test).

Referring to fig. 16A and 16B, a subset of patients (n 15) were examined in a study related to the present invention and had multiple judgments before and after electroencephalography and found a relatively stable revised coma recovery scale score at both the group and individual levels. Repetitive behavior measurements (revised coma recovery scale) may be important to accurate judgment and prediction, and the predictive value of a system comparing multiple electroencephalographic measurements to multiple revised coma recovery scale scores remains to be explored.

However, current research also has some limitations that require attention. First is language stimulation synthesized with software. Previous studies have demonstrated that personalized and meaningful stimuli can elicit more robust and reliable responses in brain-injured patients. Thus, future work may use more personalized language stimuli, e.g., designed for topics with which the patient is familiar. Second, word and sentence level processing may be enhanced when speech rates are reduced, especially for patients with impaired consciousness. It is also noted that the inter-trial phase-consistency signals recorded at 1,2 and 4hz for the healthy control groups were significantly reduced in the noisy hospital environment compared to the laboratory recordings. Improvements in data acquisition systems and recording conditions can provide higher brain electrical data quality. Finally, although some patients have been judged multiple times by electroencephalography and the accuracy of the model's judgment appears to be very high, there is also a possibility that over time, misdiagnosis may occur for some patients due to fluctuations in the level of consciousness. In fact, clinicians need to perform multiple behavioral assessments of consciousness in order to achieve a stable and accurate determination. For example, there is evidence that the number of assessments has a significant impact on the judgment of the degree of disturbance of consciousness within two weeks, even within one day, and that higher responsiveness in the morning than in the afternoon is found in behavioral assessments. Also, the level of consciousness as evident from the electroencephalogram may fluctuate over several days or every other day. Thus, future work may require the deployment of multiple electroencephalographic measurements from early in coma to subsequent recovery.

Aspects of the present application may be embodied entirely in hardware, entirely in software (including firmware, resident software, micro-code, etc.) or in a combination of hardware and software. The above hardware or software may be referred to as "data block," module, "" engine, "" unit, "" component, "or" system. The processor may be one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), digital signal processing devices (DAPDs), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, or a combination thereof. Furthermore, aspects of the present application may be represented as a computer product, including computer readable program code, embodied in one or more computer readable media. For example, computer-readable media may include, but are not limited to, magnetic storage devices (e.g., hard disk, floppy disk, magnetic strips … …), optical disks (e.g., Compact Disk (CD), Digital Versatile Disk (DVD) … …), smart cards, and flash memory devices (e.g., card, stick, key drive … …).

The computer readable medium may comprise a propagated data signal with the computer program code embodied therein, for example, on a baseband or as part of a carrier wave. The propagated signal may take any of a variety of forms, including electromagnetic, optical, and the like, or any suitable combination. The computer readable medium can be any computer readable medium that can communicate, propagate, or transport the program for use by or in connection with an instruction execution system, apparatus, or device. Program code on a computer readable medium may be propagated over any suitable medium, including radio, electrical cable, fiber optic cable, radio frequency signals, or the like, or any combination of the preceding.

Similarly, it should be noted that in the preceding description of embodiments of the application, various features are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure aiding in the understanding of one or more of the embodiments. This method of disclosure, however, is not intended to require more features than are expressly recited in the claims. Indeed, the embodiments may be characterized as having less than all of the features of a single embodiment disclosed above.

Also, this application uses specific language to describe embodiments of the application. Reference throughout this specification to "one embodiment," "an embodiment," and/or "some embodiments" means that a particular feature, structure, or characteristic described in connection with at least one embodiment of the present application is included in at least one embodiment of the present application. Therefore, it is emphasized and should be appreciated that two or more references to "an embodiment" or "one embodiment" or "an alternative embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, some features, structures, or characteristics of one or more embodiments of the present application may be combined as appropriate.

Numerals describing the number of components, attributes, etc. are used in some embodiments, it being understood that such numerals used in the description of the embodiments are modified in some instances by the use of the modifier "about", "approximately" or "substantially". Unless otherwise indicated, "about", "approximately" or "substantially" indicates that the number allows a variation of ± 20%. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximations that may vary depending upon the desired properties of the individual embodiments. In some embodiments, the numerical parameter should take into account the specified significant digits and employ a general digit preserving approach. Notwithstanding that the numerical ranges and parameters setting forth the broad scope of the range are approximations, in the specific examples, such numerical values are set forth as precisely as possible within the scope of the application.

Although the present application has been described with reference to the present specific embodiments, it will be recognized by those skilled in the art that the foregoing embodiments are merely illustrative of the present application and that various changes and substitutions of equivalents may be made without departing from the spirit of the application, and therefore, it is intended that all changes and modifications to the above-described embodiments that come within the spirit of the application fall within the scope of the claims of the application.

Claims

1. A system for assessing the level of an conscious disturbance, comprising:

a language generator configured to generate speech material that produces auditory stimuli to a subject;

an electroencephalograph configured to record an electroencephalogram signal of the subject; and

and the analysis device is configured to extract characteristic parameters from the electroencephalogram signals and obtain the consciousness disturbance degree of the subject according to the characteristic parameters.

2. The system of claim 1, wherein the stimulus comprises one or more of a word stimulus, and a sentence stimulus.

3. The system of claim 2, wherein the word stimulus comprises a single word that occurs at a first frequency, the word stimulus comprises a word that occurs at a second frequency, and the sentence stimulus comprises a sentence that occurs at a third frequency.

4. The system of claim 3, wherein the first frequency comprises 4 hertz.

5. The system of claim 3, wherein the second frequency comprises 2 hertz, the words comprise 2-word words, and the 2-word words comprise noun phrases.

6. The system of claim 3, wherein the third frequency comprises 1Hz, the sentences comprise 4-word sentences, and the 4-word sentences comprise noun phrases and/or motile phrases.

7. The system of claim 1, wherein the characteristic parameters include inter-trial phase consistency ITPC, calculated using the following formula:

8. The system of claim 1, wherein the characteristic parameters include A, B, C, D-four brain electrical map-based micro-state parameters including an a-state representing a probability of occurrence of the a-topographic map, a B-state representing a probability of occurrence of the B-topographic map, a C-state representing a probability of occurrence of the C-topographic map, and a D-state representing a probability of occurrence of the D-topographic map.

9. The system of claim 8, wherein the micro-state parameters further include a-P states representing probabilities of occurrence of front and rear topographical maps averaged based on an A, B topographical map and L-R states representing probabilities of occurrence of left and right topographical maps averaged based on a C, D topographical map.

10. The system of claim 9, wherein the micro-state parameters further comprise a difference between the a-P state and the L-R state.

11. The system of claim 9, wherein the micro-state parameters further comprise a probability weighted spatial correlation coefficient difference ac between the a-P state and the L-R state_ρThe following formula is used for calculation:

wherein n represents the total number of recording channels of the electroencephalogram recording module, I represents the ith electroencephalogram electrode, and I_iShowing the voltage values measured on the kth topographic map, A-P showing front and rear topographic maps averaged based on the A, B topographic map, L-R showing left and right topographic maps averaged based on the C, D topographic map, V_APRepresenting the voltage value, V, measured on said front and rear topographic maps_LRRepresenting the voltage values, p, measured on said left and right topographic maps_kThe appearance probability of the kth topographic map is shown, k represents the number of the topographic map, and the numbers k of the topographic map A, B, C, D are 1,2, 3, and 4, respectively.

12. The system of claim 11, wherein the micro-state parameters further include a frequency of occurrence of a-P states, a duration of L-R states, an inter-a-P-state transition rate, and an inter-L-R-state transition rate.

13. The system of claim 12, wherein the analysis device comprises a linear discriminant analysis classifier, an input of the linear discriminant analysis classifier is the feature parameter, and an output of the linear discriminant analysis classifier is a degree of disturbance of consciousness of the subject.

14. The system of claim 1, wherein the degree of disturbance of consciousness in the subject includes a state of micro-consciousness and unresponsive arousal syndrome.

15. The system of claim 1, further comprising:

a display configured to generate a visual task that produces visual stimuli to the subject;

a subject feedback button configured to receive a press by the subject, the subject pressing the subject feedback button according to the visual task.

16. The system of claim 12, further comprising a prediction device, the characteristic parameters further comprising an inter-trial phase consistency ITPC, the input to the prediction device being one or more of the inter-trial phase consistency and the micro-status parameter, the output of the prediction device being a positive result or a negative result, the positive result indicating that the disorder of consciousness of the subject has a tendency to recover, the negative result indicating that the disorder of consciousness of the subject does not have a tendency to recover.

17. The system of claim 16, wherein when the stimulus speech is a word stimulus speech, the input of the prediction means comprises: a difference between the A-P state and the L-R state in the micro-state parameters, and a frequency of occurrence of the A-P state;

when the stimulation speech is word stimulation speech, the input of the prediction device comprises: the frequency of occurrence of the A-P states in the micro-state parameters and the inter-trial phase consistency; and

when the stimulation speech is sentence stimulation speech, the input of the prediction device includes: probability-weighted spatial correlation coefficient differences between the A-P states and the L-R states in the micro-state parameters, and the frequency of occurrence of the A-P states.

18. A method for predicting a tendency to recover from an consciousness disorder, comprising:

applying a stimulus to the subject;

recording an electroencephalogram signal of the subject; and

extracting characteristic parameters from the electroencephalogram signals, and predicting whether the consciousness disorder of the subject has a recovery tendency or not according to the characteristic parameters.

19. The prediction method of claim 18, wherein the characteristic parameters comprise one or more of inter-trial phase consistency and micro-state parameters.

20. A storage medium having computer program code stored thereon, the computer program code performing the following steps when executed by a processor:

controlling a speech generator to generate speech material that produces auditory stimuli to a subject;

controlling an electroencephalograph meter to record an electroencephalogram signal of the subject; and

and the control analysis device extracts characteristic parameters from the electroencephalogram signals, obtains the degree of the conscious disturbance of the subject according to the characteristic parameters and predicts whether the conscious disturbance of the subject has a recovery tendency.