Automatic recognition of the early stage of alzheimer’s disease based on discrete wavelet transform and reduced deep convolutional neural network

Abstract

In this paper, the classification of normal controls (NC), very mild cognitive impairment and mild cognitive impairment (MCI) from structural magnetic resonance imaging (MRI) are proposed, based on the discrete wavelet transform (DWT) and reduced deep convolutional neural network (RDCNN). Multi-resolution analysis using DWT is applied to the digital images for decomposition purposes. The automatic feature extraction, selection and optimization are performed using the proposed RDCNN. The classification accuracy and learning speed of the DWT-RDCNN method are compared with RDCNN by taking the MRI data as input. The superior classification accuracy of the proposed DWT-RDCNN method over RDCNN method as well as other recently introduced prevalent methods is the major advantage for analyzing the biomedical images in the field of health care.