Research And Implementation Of Hippocampal Radiomics Based Assisted Diagnosis Model For Alzheimer’s Disease

Alzheimer’s disease is a neurologically irreversible disease and will become more severe over time.There are three medical stages in the development of Alzheimer’s disease,namely cognitive normal(CN),mild cognitive impairment(MCI)and confirmed Alzheimer’s disease(AD).Currently,the volume and morphological features of the hippocampal region in magnetic resonance imaging(MRI)have been used as important biomarkers for the diagnosis of Alzheimer’s disease;therefore,studies targeting the hippocampus are of great significance for the ancillary diagnosis of Alzheimer’s disease.This paper takes structural magnetic resonance(sMRI)images as the research object,and firstly proposes 3DUnet-CBAM model to achieve accurate segmentation of 3D hippocampal images,then extracts the image histological features of hippocampus,and finally after feature screening and fusion,training,the classification prediction of Alzheimer’s disease can be achieved.The main contents of this paper are as follows:1)In view of the small size of the hippocampus and the difficulty of accurate edge segmentation,this paper proposes a segmentation model of the hippocampus based on3DUnet-CBAM.Firstly,the data is preprocessed.The model uses 3DUnet as the basic network structure,uses 3D convolution network to avoid the information loss between twodimensional slices,and fuses CBAM attention mechanism in the last layer of down sampling,so as to effectively combine the deep and shallow features of hippocampus,and finally improves the segmentation accuracy of hippocampus.The experimental results show that the dice similarity coefficient,precision and recall rate of the model under the verification set are 89.01%,89.04% and 88.97% respectively.2)Taking the hippocampus as the research object,the concept of absolute volume and relative volume is introduced.The experimental results show that the classification performance of relative volume of hippocampus to ad is better than that of hippocampus.Aiming at different stages of Alzheimer’s diagnosis,a classification model of Alzheimer’s disease based on Imaging Group is proposed.The imaging characteristics of hippocampus are extracted first,Then,the recursive feature elimination method was used to screen the features,then the imaging group features were combined with the relative volume of hippocampus.The Ada Boost classifier was used to train.The experimental results showed that the F1 scores of AD,MCI and CN were 0.89,0.86 and 0.91.3)In this paper,combined with the above models and methods,we design and implement an auxiliary diagnosis system for Alzheimer’s disease based on hippocampal imaging omics,which includes the format conversion from two-dimensional DICOM image to threedimensional NIFTI image,hippocampal volume measurement and intelligent auxiliary diagnosis.The user can automatically segment the hippocampus by inputting the sMRI image,output the relative volume of the hippocampus and predict the disease condition.