Automatic Alzheimer Screening on Clock Drawing Test Using Deep Learning : ASDL

Alzheimer’s disease is a progressive neurological condition that impairs memory, cognitive abilities, and daily functioning, making early detection critical for effective intervention. The Clock Drawing Test (CDT) is commonly used to screening between Alzheimer’s disease (AD) and healthy control (HC) cases. However, traditional scoring methods rely heavily on subjective expert judgment, which introduces bias and variability. To address these issues, this research explores an automated deep learning-based approach for analyzing CDT images. We compare the performance of four state-of-the-art Convolutional Neural Network (CNN) models—DenseNet, ResNet, EfficientNet, and Vision Transformer (ViT)—for classifying Alzheimer’s severity using a collected dataset of CDT images. From the experimental results, DenseNet-121 outperformed the other models, achieving a training accuracy of 99.95% and a validation accuracy of 98.84%. Moreover, this model demonstrated strong performance in testing, as measured by precision, recall, F1-score, and confusion matrix. The model also revealed significant detection areas, focusing on the clock’s center and the direction of its hands, using Score-CAM interpretation. This research proposes DenseNet-121 as an effective tool for analyzing CDT images, contributing to early detection of Alzheimer’s disease in healthcare applications.