Author : Rahul P. Mahajan

It is critical to detect Alzheimer’s disease (AD) early on so that medications may be started quickly since it is a major cause of dementia in the elderly. Furthermore, a substantial portion of the world’s population is impacted by metabolic illnesses such as diabetes and AD. The progressive nature of Alzheimer’s disease makes it difficult for patients and their loved ones to plan for the future, but progress toward therapies that might halt the disease’s course is being made possible. This article showcases a machine learning-based approach to early AD detection using the Open Access Series of Imaging Studies (OASIS) dataset. Comprehensive data pretreatment is part of the technique, which also includes feature selection, data cleaning, addressing missing values, and balancing utilizing the Synthetic Minority Oversampling Technique (SMOTE). The CNN model obtained from deep learning served as the basis for comparison against XGBoost and Logistic Regression and Random Forest (RF) AD methods. Accuracy, precision, recall, and F1-score are some of the measures used to train and assess CNN. According to experimental results, the CNN model showed better performance than classical models through accuracy of 94.1% and, precision of 98%, re-call of 91%, resulting in an F1-score of 94%. This demonstrates CNN’s promise for dependable and accurate early diagnosis of AD.

Keywords: Alzheimer’s Disease, Early Detection, Machine Learning, Convolutional Neural Networks, OASIS Dataset, SMOTE.