ScholarWorks@성균관대학교

초록

Pathological brain atrophy may arise from various neurological and systemic conditions, such as depressive disorders. In response to the clinical need for auxiliary diagnostic tools to detect brain atrophy in young individuals with depression, this study proposes a deep learning-based framework for the classification of cerebral atrophy using computed tomography (CT) images. The proposed approach leverages both convolutional neural networks (CNNs) to capture spatial structural changes of the brain. It also uses multilayer perceptron (MLP) to incorporate psychological risk factors to enhance diagnostic accuracy with a visual explanation for the classification. This integration allows for a more comprehensive representation of the patient profile. Experimental results demonstrate that the CNN model achieves an accuracy of 83.01 %, while the multimodal model, which combines CT imaging data with psychological information, attains an improved accuracy of mathbf{87.25 %}. These findings underscore the effectiveness of multimodal data in cerebral atrophy classification, particularly in the context of depressive disorders. The visual explanation provides a heatmap to highlight key features influencing brain atrophy classification.

키워드