2026-05-04T18:33:20.671Z http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai

ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai:2:11788 2026-02-03T16:20:17Z 2

Metaheuristic-Driven Dual-Layer Model for Classifying Alzheimer's Disease Stages 2026 http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai/record/2/11788 https://www.frontiersin.org/journals/computational-neuroscience/articles/10.3389/fncom.2026.1731812/abstract L. Anicin S. Andjelic M. Markovic Blagojevic D. Bulaja M. Zivkovic T. Zivkovic M. Antonijevic N. Bacanin Accurate determination of the progression phase of Alzheimer's disease (AD) plays a key role in making timely medical decisions, improving patient care, and tailoring therapeutic interventions. This study addresses the challenge of assessing AD stages by introducing an advanced machine learning architecture designed to leverage complex neuroimaging data. Specifically, it proposes a composite model that combines convolutional neural networks (CNNs) for deep-level feature derivation with powerful ensemble learners, XGBoost and LightGBM, whose performance is further refined through metaheuristic optimization strategies. The framework employs a two-tier design: the first stage concentrates on extracting distinctive feature patterns from MRI images, whereas the second stage utilizes ensemble learning to achieve accurate stage prediction. Extensive experiments performed on a publicly available AD dataset demonstrate the system's strong ability to manage a multi-class classification problem spanning multiple disease phases. Across three different experimental configurations, optimized models achieved a maximum classification accuracy of 89.55%, underscoring the effectiveness of the proposed approach. In addition, explainable AI (XAI) methodologies were applied to interpret the generated results, offering clinically relevant explanations of feature importance. These findings contribute to improving future data collection strategies and support the advancement of more transparent, interpretable, and trustworthy diagnostic frameworks for Alzheimer's disease. article 10.3389/fncom.2026.1731812 20 1731812 1662-5188 Frontiers in Computational Neuroscience