Taylor-dingo optimized RP-net for segmentation toward Alzheimer's disease detection and classification using deep learning.

Alzheimer's Disease (AD) is a significant cause of mortality in elderly people. The diagnosing and classification of AD using conventional manual operation is a challenging issue. Here, a novel scheme, namely Recurrent Prototypical Network with Taylor Dingo Optimizer- enabled Deep Neuro Fuzzy Network (RP-Net_TaylorDOX-based DNFN) is devised for classifying the AD from input image. Here, the brain image is considered as the input for AD severity classification, and pre-processing is then processed by a median filter and Region of Interest (RoI). Later, segmentation of the ROI extracted image is done by RP-Net. Here, the parameters of RP-Net are tuned using the Taylor Dingo Optimizer (TaylorDOX), which is developed by the integration of the Taylor series with the Dingo Optimizer (DOX). Once finishing segmentation is performed, feature extraction is effectuated to mine the important features. After that, data augmentation is accomplished based on the oversampling technique. Moreover, AD detection is accomplished by utilizing a Deep Convolution Neural Network (DCNN), which is trained by the TaylorDOX. Finally, severity classification is effectuated utilizing DNFN structurally optimized by TaylorDOX. Furthermore, the effectiveness of RP-Net_TaylorDOX-based DNFN is calculated depending on the segmentation accuracy, accuracy, sensitivity, and specificity and has obtained values of 92.13 %, 92.30 %, 93.62 %, and 91.52 % correspondingly.