Early Alzheimer's Disease Detection with Multiple Machine Learning Models

Abstract Alzheimer's disease (AD) arises from aberrant protein buildup in the brain and impairs cognition. AD is more susceptible to therapy early on, therefore early detection is crucial for effective treatments. In the last decade, machine learning (ML) methods have successfully detected AD and other medical imaging applications. These methods can automatically learn and recover features from large datasets, making them useful for medical picture analysis. Logistic Regression, Radial Basis Function Support Vector, Decision Tree, Random forest, Adaptive-Boost, EXtreme Gradient Boosting (XG-Boost), Voting Classifier, K-Nearest Neighbour (KNN), Stochastic Gradient Descent (SGD), Quadratic Discriminant Analysis (QDA), Gaussian Naive Bayes, Multi-layer Perceptron (MLP), Extra-Gaussian Naive Bayes, and For AD predictions, ML models are assessed using accuracy, precision, recall, and f1 score using the open access series of imaging studies (OASIS) dataset. The study shows that ML models may be used to generate clinically meaningful AD diagnosis methods in MRI pictures. This paper ranks selected ML models by accuracy scores. Random forest, Voting, and Boosting classifiers achieved 100% accuracy and excellent results, while Passive Aggressive and KNN classifiers had lower scores.