Quantitative Analysis Of Multimodal Magnetic Resonance Imaging Of Drug Abusers Based On Deep Learning

Drug addiction is a chronic functional brain disease characterized by withdrawal and relapse,it has become one of the main risk factors that seriously affect human health and social stability.For the prevention of drug addiction,especially the prevention of relapse,there is still no effective method.Although treatment through obedience and Drug addiction is a chronic functional brain disease characterized by withdrawal and relapse,it has become one of the main risk factors that seriously affect human health and social stability.For the prevention of drug addiction,especially the prevention of relapse,there is still no effective method.Although treatment through withdrawal and transcranial magnetic stimulation treatment,drug dependence symptoms of drug addicts can be reduced or even completely disappeared,but under the induction of stress events,drug-related clues or drugs themselves,the original addiction memory can be reawakened.Therefore,studying the neural mechanism of drug addiction is of great significance to the prevention and treatment of drug addiction.With the rapid development of magnetic resonance imaging technology,a variety of imaging sequences such as structural imaging,functional imaging,magnetic susceptibility imaging,and perfusion imaging can be acquired.Therefore,it is of great value to explore the mechanism of drug addiction by studying the difference between multimodal magnetic resonance images of addicted brain and normal brain.In view of the fact that brain functional imaging has matured in the research of drug addiction,this work intends to analyze drug addiction by combining new quantitative susceptibility imaging technology and structural imaging.The main research contents are as follows:(1)Quantitative susceptibility mapping(QSM)was used to study the abnormal iron content in the brain tissue of drug addicted patients.First,segment and register the brain images of normal people and drug users,and then use the Laplacian method to unwrap the acquired phase images,and use the v-sharp(variable-kernel sophisticated harmonic artifact reduction for phase data)method to remove the background phase,finally,a QSM image is reconstructed based on the LSQR algorithm.By comparing the QSM values of the addicted brains and normal brains in several brain regions,the effects of drugs on iron deposition in brain tissue were analyzed.(2)In order to deal with the influence of preprocessing steps in traditional QSM reconstruction algorithms,such as image segmentation,phase unwrapping,background phase removal,and optimization regularizations,an end-to-end QSM image reconstruction network named QSM-mapping-net was proposed.The original phase images of 73 human brains was used as the input and the corresponding susceptibility maps was used as the label to train the network.The model was verified on the QSM Challenge public dataset and the drug user dataset,the experimental results illustrated that the model proposed in this paper has superiority on QSM reconstruction.Meanwhile,using QSM-mapping-net to further study the brain iron deposition in drug addicted patients,it can accurately find abnormal brain iron deposition in multiple brains.(3)Considering the shortcomings of traditional methods for detecting drug users,the detection accuracy is often affected by the detection time and the duration of drug use.In addition,the traditional detection method is too cumbersome,and it is also a waste of resources for the detection reagent.The blood detection and urine detection methods bring constant and privacy problems to the subject.Therefore,this paper proposes a SEPIRnet network model,which uses QSM images and structure MR images to train the network,which can accurately detect drug users,thereby solving the shortcomings of traditional methods.