The Study Of Computational Neural Inbrain Cognition And Brain Diseases

In recent years,with the rapid development of brain science,neuroscience and information science,all countries have launched “brain plans” in succession.The “brain plan” of China has also been basically completed.The purpose is to study the neurological principles of brain cognition as the main body to study new means of diagnosis,treatment of major brain diseases and new technologies of brain intelligence.Computational neuroscience can explain the cognitive function of the brain from a computational perspective and provide a means of researching for major brain diseases.Therefore,the study of brain cognition and brain diseases based on computational neuroscience has become an important issue.Then this thesis uses the computational methods of microstates,Lempel-Ziv complexity,common spatial pattern and sparse representation-based classification to study brain cognition and brain diseases form the following three aspects:(1)The dynamic EEG microstates in the cognitive process of mental rotation were studied.In this thesis,we mainly proposed the microstates to examine the encoding of mental rotation from the spatial-temporal changes of EEG signals.We collected EEG data from 11 subjects in a mental rotation cognitive task using 12 different stimulus pictures.We applied the classical K-means clustering method to investigate the microstates conveyed by the event-related potential extracted from EEG data during mental rotation,and obtained four microstate modes(A,B,C,D).Subsequently,we defined several measures,including microstate sequences,topographical map,hemispheric lateralization,and duration of microstate,to characterize the dynamics of microstates during mental rotation.We observed that 1)the microstates sequence had a specified progressing mode,i.e.,A?B?A;2)the microstate of the mode A demonstrated a clear hemispheric lateralization;3)the duration of microstate mode A showed a significant "angle effect".(2)EEG feature of stroke patients based on Lempel-Ziv complexity were studied.We used the LZC analysis to study brain activity of stroke patients with mental rotation task.From the experimental results,the non-linear complexity of the brain topographical map better reflected to the lesion area of the stroke patients.What's more,the nonlinear dynamic complexity of the lesion area in stroke patients is generally lower than that in normal subjects.(3)A method of identifying stroke based on common spatial pattern(CSP)algorithm and sparse representation-based classification(SRC)was proposed.To identify stroke patients and normal controls during mental rotation task,CSP was employed to extract features from EEG and then SRC was used to classify?A series of experiments demonstrated the effectiveness of features extracted by CSP of both classes and the SRC could obtain excellent results in classification.The classification accuracy of SRC was about 90%,and the classification accuracy was higher than 95% under the visual stimulus of S4 and S10(180o).Through the application in brain cognition and brain diseases,it is shown that using the methods of computational neuroscience can provide a new effective way for analyzing the cognitive process of mental rotation,and provide a basis for the early diagnosis of stroke,and lay the foundation for exploring effective stroke treatment methods in the future.