Fusion of bio-inspired optimization and machine learning for Alzheimer's biomarker analysis.

Identification of Alzheimer's Disease (AD), especially in its early phases, presents significant challenges due to the nonexistence of reliable biomarkers and effective treatments. Clinical trials for AD medications also suffer from high failure rates. Accurate diagnosis, prognosis determination, progression monitoring, and treatment effect assessment depend heavily on analysing various brain regions, including the Corpus Callosum (CC), Grey Matter (GM), Hippocampus (HC), Ventricle, and White Matter (WM). Among these, the Hippocampus plays a pivotal role in early detection. This study employs deep learning for classification and optimization techniques for segmenting the HC region to enable the AD diagnosis. The pre-processing of raw images involves histogram equalization and Otsu's thresholding methods. The research focuses on data collection and pre-processing as essential steps for advancing diagnostic methods. Segmentation and classification utilize Elephant Herding Optimization (EHO) and Crow Search Optimization (CSO) techniques in combination with the ResNet50 classifier. The results reveal that Crow Search Optimization achieves superior performance, with an accuracy of 92 %, surpassing Elephant Herding Optimization.