Suraksha Vinod, Sushil Bohara, Catherine Lacey
Over 6.7 million individuals over 65 years old in the United States are reported to have been affected by Alzheimer’s Disease (AD), a neurodegenerative disorder that is characterized by the accumulation of amyloid plaques and tau tangles in the brain. These changes lead to structural and functional changes in the brain gradually resulting in the worsening of cognitive abilities, memory loss, and impaired functioning in daily life. Currently, AD remains hard to clinically diagnose as it relies on observable symptoms, which may not be present until 10 years after damage begins due to the disease. This delay in diagnosis, coupled with the lack of a cure, poses a significant obstacle to early intervention, which is crucial in order to slow the disease progression. In this study, we aim to leverage computational approaches, specifically machine learning, to analyze functional Magnetic Resonance Imaging (fMRI) data from Alzheimer’s Disease patients, Mild Cognitive Impairment (MCI) patients, and control groups. We evaluated white box and black box SVM, RF, and CNN (only black box) machine learning models to assess their accuracy in classifying and thus predicting AD. The white-box RF and black-box CNN models performed the best with accuracies of 78% and 77.7%, respectively, but the RF offers interpretability making it the best ML model for AD classification. Our findings demonstrate the potential of machine learning to significantly enhance diagnostic precision, support earlier clinical interventions, and ultimately contribute to slowing disease progression and improving patient outcomes.
Available on request
Obtained from the ADNI database
Simply open the Jupyter Notebook and run the cells. Our processed data is available on request. Feel free to modify the paths and code to suit your data needs.