Research On Automatic Identification Of Alzheimer's Disease And Epilepsy Based On MRI

Alzheimer's disease(AD)and epilepsy are common chronic diseases in neurology,and their pathogenesis is complicated,which not only seriously harms the physical and mental health of patients,but also brings a heavy burden on families and society.The two have some correlation in etiology and pathological changes.With the rise of machine learning technology,more and more automatic diagnostic technologies have been proposed one after another.This technology can provide doctors with objective,consistent and reproducible reference information for correct intervention and treatment of patients.Magnetic resonance imaging(MRI),as a commonly used brain clinical examination method,plays an important role in the diagnosis and treatment of AD and epilepsy.The development of automatic MRI analysis methods for these two diseases has very important practical significance.This thesis has carried out three studies on technologies related to MRI automatic identification of AD and epilepsy.The main contents and achievements include:1.A neural network architecture that fuses MRI,clinical and genetic data is proposed to predict whether patients with mild cognitive impairment have a high risk of AD conversion.Considering that some clinical and genetic factors may affect this conversion,in addition to MRI data,clinical data and genetic data are also included in the study.A 3D convolutional neural network is constructed to process MRI data,and a multi-modal data fusion strategy is applied to combine information from various sources.The introduction of clinical data and genetic data has improved the prediction effect of AD conversion,indicating that this study has potential and value in the development of auxiliary diagnostic systems for AD prognosis.2.A 3D convolutional neural network architecture using multi-modal MRI is proposed,for the segmentation of three types of brain tissue: gray matter,white matter,and cerebrospinal fluid.This network relies on the semantic segmentation idea of the full convolutional network,based on the encoder-decoder structure,takes multi-modal MRI data as input,and combines residual units and dilated convolutions.The introduction of multi-modal MRI and dilated convolutions has significantly improved various indicators,indicating that this study can be applied to the development of automatic brain tissue segmentation systems.3.A method for MRI automatic identification of both AD and epilepsy is proposed.This method is based on the idea of voxel-based morphometry and consists of three steps: preprocessing,feature extraction and classification.The selection of brain tissue categories and feature extraction methods are studied to find the most suitable combination for experimental samples.At present,there are few automatic identification systems for epilepsy,and this study can provide a reference for the development of related auxiliary diagnostic systems.