Characteristics Of Oscillation Frequency In Resting-state BOLD Signal And Their Clinical Application

The human brain is a sophisticated and complex system whose internal neural process is usually a combination of multiple neural signals with different frequencies.Previous studies have shown that low-frequency oscillation signals in brain neural activity are closely related to many advanced cognitive functions,and low-frequency oscillation signals are also important for revealing brain functional abnormalities in some diseases.With the development of medical imaging technology,the blood oxygenation level dependent(BOLD)oscillation provides an effective approach to investigate the low-frequency oscillation characteristics of neural activity.However,there are three main problems in previous studies.Firstly,there is no reliable and unified method on dividing BOLD oscillation frequency.Secondly,it lacks systematical study on the effects of BOLD oscillation on whole-brain functional connectivity,especially on dynamic functional connectivity.Thirdly,it is still unclear the efficiency of BOLD oscillation features in discriminating neuropsychiatric patients from healthy controls.In this thesis,the frequency component analysis method is constructed to systematically study the influence of the BOLD oscillation frequency on the brain functional connectivity as well as the effects on diagnosing neuropsychiatric diseases.The research consists of the following three parts:In the first part,the frequency component analysis method was constructed.Based on the complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)method and the Hilbert-Huang transform(HHT),the frequency component analysis was constructed.For one thing,this method can adaptively decompose the BOLD oscillation frequency into different sub-bands,and for another,the frequency component characteristics of BOLD oscillation can be evaluated at the whole-brain level.In the second part,the thesis systematically studied the influence of the BOLDoscillation frequency on whole-brain functional connectivity.Based on the frequency component analysis,the frequency-dependent changes of whole-brain functional connectivity in neuropsychiatric diseases were systematically investigated from static and dynamic aspects.On the one hand,functional connectivity density(FCD)and betweenness centrality(BC)were used to examine the characteristics of static local functional connectivity in the whole-brain of neuropsychiatric diseases in different frequency bands from voxel levels and large-scale network levels.The results indicated that the brain region with abnormal local efficiency in patients with generalized anxiety disorder(GAD)has frequency specificity,and the FCD value of local efficiency change in specific frequency bands can be used as a potential biomarker for GAD.Meanwhile,at the large-scale network level,the BC value of the default mode network in the seizure interval and that of the internal central node in the sensory/somatomotor network were significantly correlated with the severity of the disease at specific frequencies of 0.015-0.025 Hz and 0.12-0.25 Hz,indicating that frequency-specific changes of local efficiency in the functional connectivity network can be used as a reference for disease progression.On the other hand,the time-varying connectivity analysis was used to study the frequency-dependent features of the dynamic functional network connectivity.Dynamic functional connectivity(DFC)analysis which is considered the frontier in the field of computational neuroscience can effectively describe the time-varying properties of brain functional activity.In this thesis,the DFC of the resting-state subnetwork of epilepsy patients in different frequency bands was investigated for the first time by constructing dynamics evaluation indicators,combined with sliding window and cluster analysis methods.The results showed that the abnormal changes in DFC between the resting-state sub-network of patients with epilepsy also have frequency-dependent characteristics,and the dynamics abnormality of subcortical network in the frequency band of 0.0095-0.0195 Hz is of important significance for diagnosing specific symptoms of epilepsy.In the third part,the effect of BOLD oscillation frequency in the diagnosis of neuropsychiatric diseases was studied.By extracting brain function image features and combining machine learning methods,the identification of neuropsychiatric diseases is a hot research topic in the field of medical aided diagnosis.However,it is unclear whether the BOLD oscillation frequency characteristics can be used as a classification feature to identify neuropsychiatric diseases.Therefore,this thesisextracted FCD values at different frequencies as classification features,and investigated the effects of frequency-specific local functional connectivity on discriminating patients with GAD from healthy controls(HCs).The results showed that the frequency-specific FCD value can effectively discriminate patients with GAD from HCs(accuracy rate 89.83%),and the insula has the highest classification weight in the lower frequency range of 0.02-0.036 Hz,indicating that the frequency-specific local functional connectivity of the insula can be used as an effective biomarker for GAD diagnosis.Furthermore,based on the frequency component analysis,the weighted value of frequency component was calculated,and the classification characteristics were used to discriminate patients with depression and HCs.The results indicated that the weighted frequency characteristics of BOLD oscillation can effectively distinguish patients with depression from HCs(accuracy rate 87.05%),and the ratio of low-frequency components(0.01-0.08 Hz)in BOLD oscillation is of great significance to identify depression.Meanwhile,the weighted frequency characteristics of the cogniton connectivity network related brain regions,especially the middle frontal cortex,showed higher weight in the classification,suggesting that the frequency specificity of the brain region can be used as a clinically effective biomarker for depression.In summary,this thesis reveals that BOLD oscillation frequency not only affects the local efficiency of the static functional connectivity,but also affects the time-varying properties of the dynamic functional connectivity.Some specific frequency ranges are more sensitive in detecting abnormalities of brain function in neuropsychiatric diseases.In addition,this thesis also indicates that BOLD oscillation frequency characteristics have a good classification efficiency for discriminating patients with neuropsychiatric diseases from HCs,which is of an important theoretical significance and application vale for using BOLD oscillation frequency characteristics in the aided diagnosis of neuropsychiatric diseases.