Multi -Class Classification of Neurodegenerative Diseases from Brain MRI Using Deep and Traditional Machine Learning Models

Título: Multi -Class Classification of Neurodegenerative Diseases from Brain MRI Using Deep and Traditional Machine Learning Models

Autores: Lucas Silveira, Rogerio Martins Gomes

Resumo: Neurodegenerative diseases are a growing global health concern, often leading to progressive cognitive and motor impairments. Accurate and early diagnosis remains a challenge due to the complexity and overlap of clinical symptoms. This study proposes an automated classification approach for Alzheimer’s disease, Parkinson’s disease, Lewy body dementia, and stroke using structural brain magnetic resonance imaging (MRI). We compared the performance of traditional machine learning algorithms—Random Forest, Support Vector Machines (SVM), and XGBoost—against a Convolutional Neural Network (CNN) in a multi-label classification framework. The models were trained and evaluated using over 18,000 MRI slices extracted from the OASIS-3 database. Our best model, a CNN architecture, achieved 95.07% accuracy and a Macro F1-score of 84.62% in a fixed-label prediction scenario. Additionally, we implemented a computer-aided diagnosis (CAD) system that enables real-time image visualization and probabilistic diagnostic output. The results highlight the potential of AI-based tools to support clinical decision-making in the early detection of neurodegenerative diseases.

Palavras-chave: Machine learning; neurodegenerative disease; MRI; classification; CNN; computer-aided diagnosis.

Páginas: 8

Código DOI: 10.21528/CBIC2025-1175593

Artigo em PDF: CBIC_2025_paper1175593.pdf

Arquivo BibTeX:
CBIC_2025_1175593.bib