Early Diagnosis Model And Medical Data Analysis Of NPSLE Based On Deep Learning

Neuropsychiatric systemic lupus erythematosus(NPSLE)is a manifestation of systemic lupus erythematosus(SLE)involving the nervous system.It has a high morbidity and mortality.It is one of the main causes of death and poor prognosis factors in SLE patients.At present,magnetic resonance spectroscopy(MRS)is mainly used to analyze the changes of metabolites in different brain regions of patients with SLE.MRS can noninvasively detect the changes of metabolites in different brain regions of SLE patients,and provide the basis for the early diagnosis of NPSLE patients.The traditional MRS acquisition is related to many factors,so it often affects the objectivity and robustness of the measurement results of intracranial metabolites.Therefore,based on the changes of intracranial metabolites in patients with NPSLE,and different from the traditional single voxel MRS,which can only detect metabolic changes in one brain region,multi voxel proton magnetic resonance spectroscopy(MVS)was used to simultaneously detect 13 kinds of intracranial metabolic changes in9 brain regions of the experimental samples.At the same time,an early diagnosis model of NPSLE based on deep learning technology is proposed.In this model,the recursive feature elimination algorithm based on SVM(SVM-RFE)is used to reduce the dimension of metabolic features,and the most relevant metabolites are selected for better recognition of NPSLE.In order to accurately judge the changes of intracranial metabolites in NPSLE patients,this paper constructs a Deep Stacking Network based on SVM(SVM-DSN)to quantitatively analyze the selected features,so as to achieve the effect of accurate diagnosis of NPSLE patients through the changes of intracranial metabolites.The results show that although the changes of brain metabolites in NPSLE are very subtle and can hardly be judged by vision,the diagnostic model can still accurately distinguish NPSLE patients and normal controls.This study is not only helpful for the early diagnosis and intervention of NPSLE,but also can reduce the bias of artificial screening.It also provides a new idea for intelligent assistant diagnosis of other diseases.In addition,in order to better make medical diagnosis serve medical health and promote the combination of science and technology and clinical practice,this paper constructs a medical data analysis platform,starting from the data collection,data cleaning and data analysis of medical data,and finally realizes the visualization of medical data.Taking SLE as an example,the data of patients,diagnosis results and treatment process are imported into the data analysis platform to realize the sharing and analysis of medical data.The usability and interpretability of medical data are enhanced through visualization and other technical methods.