MRI Brain Image Classification Based on Improved Topographic Sparse Coding

摘要

As the field of neuroimaging grows, imaging technology plays an increasingly important role in the auxiliary diagnosis of brain lesions. An Alzheimer's disease recognition method for magnetic resonance imaging (MRI) based on improved topological sparse coding (ITSC) is proposed. The data source is taken from the ADNI database, after correction, registration, segmentation, smoothing and other operations, the gray matter image of the brain are obtained, and then the improved topological sparse coding model of unsupervised feature learning is used to construct the deep neural network. And the optimization algorithm is used to replace the network model. The valence function is optimized iteratively, the weight matrix is studied and the new feature expression is obtained. Finally, the Softmax classifier and the auxiliary fine-tuning method are used to identify the disease. Compared with principal component analysis and a self-learning neural network, the experimental results show that the proposed method has better recognition performance.