Brain Functional Connectivity In Drug Addicts Based On Resting-state Functional Magnetic Resonance Imaging

Drug addiction(DA)is a chronic brain disease.Long-term drug use leads to changes in the functioning of some brain structures resulting in compulsive relapse to drug use.Therefore,exploring the causes of relapse to drug use is of clinically important significance for abstinence.Functional magnetic resonance imaging(f MRI),a kind of non-invasive imaging technology,has been widely used in the studies of brain diseases because it not only occupies the high-resolution property of magnetic resonance imaging,but also can effectively detect brain activity.Currently,most seed point-based methods focus on the brain functional connections based on regions of interest(ROI),which would ignore some important connections due to the complexity of brain functional connections.Therefore,this work tends to explore the differences in brain functional connections between the drug addicts and the healthy people by studying the static and dynamic functional connections of the whole brain.Moreover,we used the deep learning neural network to distinguish between drug addiction and health control(HC),which advances to find the brain areas that causing drug addiction.The main contents are as follows:(1)Combination of Functional Connectivity(FC)and Graph Theory enables analysis of message processing and passing between brain areas from multiple perspectives,which addresses the dependence between long-distance brain areas that cannot be considered for static FC analysis.The results showed that the FC of the visual network,control executive network and default mode network weakened in the DA group compared to the HC groups using the static FC method,while differences in marginal network occurs between the DA and HC groups using the small-world network model.(2)Dynamic brain functional connectivity of the whole brain by using sliding window technique and K-means algorithm can somehow avoid establishing the dynamic from the difference results based only on the static FC which ignores the impact of other brain areas on results.The results demonstrated that the functional connectivity of the ventral attention network(prefrontal cortex),control executive network(posterior cingulate gyrus)and default mode network(hippocampus)significantly weakened in the DA group,along with that of the sensorimotor network(insula)significantly enhanced.Differences of brain structure using the dynamic FC were greater compared to those using the static FC method,which further characterizes the damage caused by drug addiction to brain areas.(3)The proposed classification model with high accuracy,which outputs the DA image markers,can address that the existing models are less explanatory.The results showed that the model not only gained similar finds as that in the works(1)and(2),but also obtained the abnormal FC in the sensorimotor network(insular),control execution network(cingulate gyrus(36,94)),and default mode network(prefrontal cortex,hippocampus)in the DA group,indicating that classification is an effective way for finding image markers.