UC Irvine

In recent years, the problem of detecting Alzheimer’s disease with computer-aided diagnosis systems has become a relevant research field for medical diagnosis efficiency and image classification innovation. Alzheimer’s disease is an irreversible progressive neurogenerative disorder caused by damage to nerve cells, which leads to memory loss and other cognitive functioning skills deterioration. Detecting Alzheimer’s in its preliminary stages is crucial to planning treatment and therefore delaying the progression of the disease. Magnetic Resonance Imaging scans can capture complex changes in the brain and assess the damage caused by the disease. The growing popularity, accuracy, and applicability of Convolutional Neural Networks make them an optimal solution to perform this medical task.

This study implements and compares several deep models and configurations, including both two-dimensional and three-dimensional convolutional neural networks. The results classifying normal cognitive subjects versus Alzheimer’s patients show a good performance, especially for AlexNet, ResNet18, and ResNet34 architectures. In particular, 3D models are able to achieve better accuracy outcomes.