Predicting Alzheimer's disease from environmental risk factors: An fMRI-based functional connectivity and advanced...

Alzheimer's disease (AD) is a prevalent and severe neurodegenerative disorder influenced by both genetic and environmental factors-such as air pollution, toxic elements, pesticides, and infectious agents. In recent years, machine learning techniques have become essential in biomedical research, advancing fields like drug delivery and medical imaging through predictive modeling and pattern recognition. Functional connectivity derived from functional magnetic resonance imaging (fMRI) serves as a promising noninvasive biomarker for AD by mapping the brain's connectome and revealing neural network disruptions. In this study, we employed the Robust Multitask Feature Extraction Method to evaluate six supervised machine learning algorithms logistic regression, naïve Bayes, support vector machine, random forest, XGBoost, and CatBoostmfor AD diagnosis. A dataset of 140 fMRI images from an equal number of AD patients and healthy individuals (mean age 67.3 ± 6.7 years) was analyzed. The XGBoost algorithm demonstrated exceptional performance, achieving an accuracy of 98.2%, a recall of 96.6%, perfect precision (100%), an F1-Score of 98.2%, and a Matthews correlation coefficient of 0.96 effectively minimizing false positives and negatives. Although CatBoost and Random Forest also yielded robust results, logistic regression and naïve Bayes showed lower reliability. Overall, XGBoost emerges as a robust solution for the early and precise prediction of Alzheimer's disease, carrying significant implications for proactive patient care and treatment strategies. Beyond these findings, emerging research is exploring multimodal imaging techniques-such as PET and EEG and deeper neural network architectures to further enhance early diagnostic accuracy and treatment personalization in AD.