Entropy-based Complexity Analysis Of Resting State FMRI Signals In Human Brain And Application

The human brain is a nonlinear complex system.The research for its complexity provides a new perspective to explore the brain,which helps to discover the neural mechanisms of brain diseases.Entropy is a commonly used method for complexity analysis,describing the disorder and chaos of non-stationary signals.It has significant advantages in analyzing finite-length noisy physiological signals.In recent years,researchers have applied entropy to the study of functional magnetic resonance imaging(fMRI)signals from the human brain,analyzing complex characteristics of the brain activity,and have achieved some results.However,there are still some key issues that need to be resolved when applying entropy to the analysis of fMRI signals.First,there is no systematic and comprehensive analysis of the consistency and stability of various entropy algorithms used in fMRI signal analysis.Second,the preprocessing cannot altogether remove the noise contained in the fMRI signal,so entropy algorithms with strong anti-noise performance are necessary.Third,single-scale entropy analysis is inadequate to reflect the different activity patterns of the brain,and single-voxel entropy analysis cannot reflect the spatial complexity among multiple voxels.Thus,entropy algorithms are urgently needed to reflect the complexity of fMRI signals more comprehensively.Therefore,this dissertation firstly analyzed the test-retest reliability of classical entropy algorithms and used entropy to study the complexity of fMRI brain signals in patients with Alzheimer's disease.Taking into account the higher stability of fuzzy entropy,but the anti-noise performance needs to be improved.Adding the sort symbolization into the fuzzy entropy,this dissertation proposed the permutation fuzzy entropy(PFE)with strong anti-noise performance.The time scale was extended to multiple time scales,and the temporal domain was extended to the spatial-temporal domain.The multiscale PFE and spatial-temporal PFE were respectively proposed.These entropy algorithms were analyzed from anti-noise performance and test-retest reliability to confirm that these algorithms were reliable for fMRI signal analysis.To explore the good application value of these improved algorithms,they were applied to the complexity analysis of fMRI signals from patients with Alzheimer's disease to find out the possible damaged brain areas.The main innovative works and results include:(1)Analyzing the test-retest reliability of classical entropy algorithms to compare the stability of classical algorithms.Approximate entropy,sample entropy,fuzzy entropy and permutation entropy have been widely used to analyze various physiological signals.The higher the test-retest reliability of entropy,the more stable the algorithm can ensure the repeatability of research results.However,there are only a few related studies.This dissertation used resting-state fMRI data sets to compare and analyze the test-retest reliability of entropy algorithms.The results showed that the test-retest reliability of permutation entropy and fuzzy entropy was higher than approximate entropy and sample entropy,which may be related to the anti-noise performance of the permutation entropy and fuzzy entropy.In addition,the test-retest reliability of permutation entropy was better than fuzzy entropy.Based on the better stability of permutation entropy,permutation entropy was used to analyze the complexity of fMRI brain signals in patients with Alzheimer's disease.It was found that the complexity of patients in the inferior temporal gyrus,middle frontal gyrus,cingulate gyrus and other regions was significantly reduced.(2)The PFE was proposed by improving the anti-noise performance of fuzzy entropy.The resting-state fMRI signal is very susceptible to noise during the acquisition process.The noise will reduce the reliability of data analysis.Preprocessing cannot completely remove the influence of noise.Therefore,the anti-noise performance of entropy is very crucial.Compared with other entropy algorithms,fuzzy entropy has better relative consistency and processing characteristics of short data sets;however,its anti-noise performance needs to be improved.This dissertation added the step of sorting symbolization into the fuzzy entropy and proposed PFE.Its anti-noise performance was analyzed through the resting-state fMRI simulation signal,and the results showed that its anti-noise performance was better than permutation entropy and fuzzy entropy.The test-retest reliability was analyzed using the resting-state fMRI data sets.The results showed that the test-retest reliability of PFE was better than permutation entropy and fuzzy entropy.When the algorithm was applied to the complexity analysis of patients with Alzheimer's disease,the complexity of the hippocampus,inferior frontal gyrus,inferior temporal gyrus,superior temporal gyrus and other regions was significantly reduced,and significantly related to the cognitive scores.(3)The multiscale PFE was proposed by expanding from a single time scale to multiple time scales.The fMRI signal of different frequency bands reflects different brain activity patterns,and single-scale entropy analysis cannot capture the complexity of signals in other frequency bands.This dissertation expanded the PFE and proposed multiscale PFE,reflecting the complexity of multiple specific frequency band signals.The simulated resting-state fMRI signal was used to analyze the anti-noise performance of the PFE on multiple time scales.The resting-state fMRI data sets were used to perform test-retest reliability analysis on multiple time scales.The results showed that the anti-noise performance and test-retest reliability of multiscale PFE were better than other multiscale entropy algorithms.When analyzing the complexity of fMRI signals in patients with Alzheimer's disease,significantly different regions were found at different time scales,reflecting that the brain has different brain activity patterns in various frequency bands.(4)The spatial-temporal PFE was proposed by expanding from the temporal domain to the spatial-temporal domain.The human brain has spatiotemporal chaos in the resting state,and the complexity analysis based on a single voxel cannot reflect the spatial complexity among multiple voxels.Combining the spatial information of the brain,expanding the temporal domain to the spatial-temporal domain,this dissertation proposed the spatial-temporal PFE.Using the resting-state fMRI signal,the anti-noise performance and test-retest reliability of the spatial-temporal PFE were analyzed and compared with other algorithms.The results showed that the temporal-spatial PFE had stronger anti-noise performance and higher test-retest reliability.Finally,it was applied to the spatiotemporal complexity analysis of fMRI signals of patients with Alzheimer's disease.It was found that the spatiotemporal complexity of some brain regions decreased significantly,such as the inferior occipital gyrus,inferior temporal gyrus,superior frontal gyrus and hippocampus.