A comprehensive hybrid model: Combining bioinspired optimization and deep learning for Alzheimer's disease identification

Abstract

Alzheimer's disease (AD) is a progressive neurodegenerative disorder characterized by a gradual decline in cognitive ability and memory function. It is a progressive disease characterized by worsening dementia symptoms over time, starting with mild memory loss and advancing to severe impairments in communication and responsiveness. The key motive of this work is to explore the new hybrid technique with parameter tuning for segmenting the brain sub regions, by satisfying the clinical targets in finding major biomarkers of Alzheimer's disease (AD). This study aims to explore a new bioinspired hybrid technique for segmenting various brain regions to assist in Alzheimer's disease (AD) diagnosis. To refine the segmentation accuracy, a combination of Gray Wolf Optimization (GWO) and Harris Hawk Optimization (HHO) approaches is proposed. Initially, the population is divided into two smaller groups, designated HHO and GWO, where both methods update subpopulation positions concurrently. The segmented region (SR) is validated with ground truth (GT) using different statistical measures and shows the highlighted accuracy of 92 %. According to the experimental results, the proposed hybrid technique provides an improved solution than HHO. After segmentation, deep learning (DL) technique is applied to classify the normal controls (NC) and AD. A 90 % classification accuracy is achieved when HHO and GWO are combined. Finally, the results are validated with clinical score, namely Mini Mental State Examination (MMSE), which gives the support to observe the disease progression. This process demonstrates the hybrid method and it proves with remarkable results, in segmentation, classification and clinical validation. In this work, clinicians have assisted in disease progression monitoring, thereby improving patient care. Significance: This work provides assistance to diagnose AD and doctors to make a decision for further treatment.

Keywords: Biomedical image processing; Clinical score; Gray wolf optimization; Harris hawk optimization; Machine learning.