Abstract: With the rapid development of human science and technology and the continuous updating of medical imaging equipment, medical imaging technology plays an increasingly important role in the auxiliary diagnosis of brain lesions. An improved L-BFGS (limited memory Broyden-Fletcher- Goldfarb-Shanno) sparse denoising self-coding network model is proposed and applied to the recognition of Alzheimer’s disease (AD) in MRI brain images. The experimental data source is taken from the ADNI data set, and the original data is corrected, registered, segmented, smoothed, etc. to obtain gray matter images of the brain. Then, the improved unsupervised greedy pre-training method and L-BFGS algorithm are combined to train the deep self-coding network as well as the learning features through Softmax regression training, so as to realize the brain mapping of patients with symptoms image recognition. ILSDAE network model is of good robustness, and compared with the stack self-coding and self-learning methods, the experimental results show the effectiveness of the proposed method.

Key words: Alzheimer’s disease (AD), L-BFGS, sparse denoising autoencoder, MRI brain image