Advances in the study of neurocharacterization of language processing at the level of patients with consciousness disorder

On May 25, Wang Liping Research Group of the Center for Excellence in Brain Science and Intelligent Technology of the Chinese Academy of Sciences (Institute of Neuroscience), Wang Liping Research Group of the Primate Neurobiology Key Laboratory of the Chinese Academy of Sciences and Mao Ying/Wu Xuehai team of neurosurgery at Huashan Hospital affiliated with Fudan University published a collaborative research paper entitled "Exploring Language Processing at the Level of PatientS with Cognitive Disorders" online in the journal Nature-NeuroscienceThis study carried out an exploratory study of neurocharacterization related to language processing in patients with cognitive impairment, and combined the relevant neural characterization with machine learning methods to successfully achievediagnosis andand rehabilitation prediction of the patient's state of consciousness, providing a new reference for the clinical diagnosis and treatment of patients with consciousness disorders (Figure 1)every year, nearly 100,000 patients in China because of brain trauma, stroke, ischemic hypoxic encephalopathy and other diseases into a coma, and then into a long-term consciousness disorder, that is, the traditional sense of the "plant man" state, long-term treatment to the family and society have brought great pressureIn patients with consciousness disorders, unresponsive wakesyndrome (Unresponsive wakeness syndrome, UWS) and minimal consciousness state (MCS) were the most common, and those with least conscious state had higher residual awareness levels and higher likelihood of recovery than those with non-reactive awakening syndromeHowever, at this stage, the patient's status identification depends on the observation and scale score of experienced doctors, which has a certain subjectiveity and a misdiagnosis rate of up to 40%In recent years, some researchers have used electroencephalography or magnetic resonance imaging to record the brain activity of patients, and then speculate on the degree of cognitive impairment, but the paradigm is more limited, neural indicators are also more scarce, accuracy still has some room for improvementTherefore, the study of language processing, which combines the advanced cognitive function of the brain with the state of consciousness, has important scientific value for the study of the brain's ability to process language processing and neural mechanism in different state of consciousness, on the other hand, it is of great clinical value and social significance for the cognitive ability of the brain as an indicator of the cognitive ability of the brain to judge the residual consciousness level of
patients, and to the diagnosis andand rehabilitation prediction of patients with cognitive impairmentAlthough researchers at this stage have not fully understood the neural basis and computer system of language hierarchy in the brain, the results of the existingbasic researchcan provide reference for clinical transformationIn previous work, the study's co-author, Ding Wei, found that when healthy people listen to Chinese speech sequences presented at a certain frequency, the brain acts at different levels of language structure (words, words, and sentences/phrases; Figure 2a) in parallel characterization sequences of neural oscillations corresponding to rhythmsFor example, when a 4-word sentence such as "pony crosses a river" is played continuously at a rate of 4 words per second, the brain tracks the words ("small," "horse," "cross," "river"; the frequency of occurrence is 4 Hz), the word ("pony," "cross the river"; the frequency of occurrence is 1 Hz) structure, and is reflected in nerve signals such as magnetic brain or brain powerCorrespondingly, if the present "High Xueshan Run" is presented with a 4-word voice sequence that does not exist in the structure of words and sentences, only nerve signals corresponding to the frequency of the word structure can be observed on this basis, the team hypothesized that the residual consciousness level of patients with consciousness disorders may be associated with the processing depth of the hierarchical structure in the language sequence, especially in the neural characterization of the high-level language structure For this reason, the researchers first designed three language sequences with different levels (single word sequences at the word level, phrase sequences of words and word structures, sentence sequences at the word, word, and sentence levels), recorded bedside brain activity in patients with non-reactive awakening syndrome and minimal consciousness, and compared them to healthy people The results of the group comparison showed that both the patient group and the healthy subject group showed significant neuroresponse to the word hierarchy, but only the cerebellum activity of the healthy subject group significantly reflected the tracking of word and sentence structure Notably, in individual analysis, the researchers found that 15 patients showed neural activity in tracking the structure of words and sentences, and six of them showed significant level seistic recovery after 100 days of cerebellum records Machine learning further shows that neural activity under the conditions of phrase sequence and sentence sequence is more effective for distinguishing between the two groups of patients in neuromechanics, consciousness is not a static brain function, but a real-time evolution of dynamic change, self-retention, and brain-wide work together Based on the relationship between consciousness and advanced cortex brain regions in the global work space theory of consciousness, the team further hypothesized that high-consciousness level brain activity would remain in the higher prefrontal-top cortex information loop for a long time, while the low-consciousness level of brain activity was more commonly distributed in low-level information processing brain regions such as sensation - that is, the more complex the language sequence synth structure of the brain, the more advanced brain region activity involved The researchers recorded the cerebellum micro-state of three groups of subjects in three language sequences with different hierarchical structures and compared them to resting states The electroencephalogram state is based on the total brain EEG electrode activity variation when the largest e-brain topographic map cluster is obtained, including duration, occurrence and other indicators, can reflect the overall activity of the brain of the temportic and spatial characteristics (Figure 2b) It was found that the differences between groups and tasks were reflected in several indicators of the four electroencephalopathy states identified by group electroencephalopathy The inter-group comparison showed that when dealing with multi-level language sequences, the health subjects showed more cerebellum micro-states associated with advanced cognitive activity information loops, while the patients had a higher probability of appearing in the cerebellum state associated with sensory information processing Comparing the two groups of patients, it was found that the duration of the sensory related micro-state in the least conscious state group was shorter than that of the non-reactive awakening syndrome group, while the advanced cognitive-related micro-state occurred more frequently in units of time than the non-reactive awakening syndrome group The inter-task comparison showed that the micro-state difference of brain activity between patients was more pronounced under advanced language task conditions (Figure 3) The results of both machine learning methods show that combining the neuroelectric activity associated with the language hierarchy with the electroencephalopathy (Figure 2c) can accurately determine the patient's level of consciousness, and the difference is best under sentence conditions more importantly, the machine learning model established by using the electroencephalogram index under the above language paradigm not only significantly better than the clinical evaluation based on behavioral scale in the of diagnosis, but also can more accurately predict the recovery of individual patients after 100 days of electroencephalography, with a correct rate of 80% In order to fully detect the reliability of language-brain model classification and prediction results for individual patients, the researchers further introduced a new batch of patient data, standardized verification of the prediction function of the current model, and successfully reproduced the accurate classification and prediction effect In contrast, models based on clinical behavioral scoring do not have the ability to generalize new data sets This series is important for the validation of the ability to predict the effects of generalization, meaning that the team may have found an evaluation of the level of general brain consciousness, which has a wide range of potential applications in a range of assessments of the state of the brain associated with consciousness levels, such as coma, sleep, anesthesia, etc The research team will further optimize the test program in future studies, combine multi-modal testing and recording techniques to further improve the clinical effectiveness, adaptation and automation of the results, and ultimately provide experimental basis and theoretical basis for the study of neural mechanisms of consciousness disorders, neural representation of consciousness activities, and related scientific research in patients with cognitive impairment (
Bioon.com of Biological Valley)