Claustrum and Hippocampus Segmentation-Based Alzheimer's Disease Identification Model Using RoT-Kmeans and CoLU-CNN

Alzheimer's disease (AD) is a dementia disease that causes loss of cognitive functions. Also, it is a noncurable disease. However, early diagnosis and proper medication reduce AD's progression time. Yet, the prevailing AD diagnosis models did not concentrate on the claustrum in the brain, reducing the efficiency of the AD diagnosis. Thus, this framework proposes an effective AD identification model with claustrum segmentation based on Collapsing Linear Unit-Convolutional Neural Network (CoLU-CNN). Primarily, the resting state-functional Magnetic Resonance Imaging (rs-fMRI) is preprocessed. Then, Gray Matter (GM), White Matter (WM), Cerebrospinal Fluid (CSF), and the hippocampus are segmented. By using Rogers and Tanimoto-based K-means (RoT-Kmeans), the putamen edge is detected from the segmented GM to segment the claustrum. Likewise, the time-series extraction is performed from rs-fMRI, and network connectivity is generated by using Seed-Based Functional Connectivity (SBFC). Then, the network connectivity and segmented claustrum are mapped. Next, the features are extracted from the segmented and mapped images, and optimal features are selected by using Kent Map-based Wild Geese Optimization (KM-WGO). Lastly, the AD is classified by using CoLU-CNN. The experimental investigation stated that the proposed methodology attained 99% AD classification accuracy.