Vision Transformer-Based Framework for Multi-Class Neurological Disease Detection Using Medical Imaging

Abstract

Neurological illnesses like Alzheimer's disease, brain tumors, multiple sclerosis, and other neurodegenerative disorders need precise and early diagnosis to be effectively treated clinically and to manage the disease. Classical diagnostic methods are often lengthy and heavily reliant on specialist interpretation of the clinical picture. To overcome these issues, this paper proposes a Vision Transformer-based design for detecting multi-class neurological diseases with state-of-the-art deep learning methods. The proposed framework applies the pretrained vit_base patch16 224 architecture, utilized to extract global, contextual, and spatial features on neurological medical images. Preprocessing of the MCND dataset involves image resizing, normalization, and transformation of tensors before training a model. Transfer learning is used to enhance prediction and predictive efficiency. Evaluation of the framework using various performance metrics is assessed in terms of accuracy, precision, recall, F1-score, confusion Matrix analysis, and ROC-AUC curves. The experimental findings show that the proposed model has stable and high classification performance on various neurological disease categories with successful learning behavior and low overfitting. The framework also exhibits high robustness, scalability, and the ability to render consistent predictive capabilities in intelligent analysis of medical images. It offers a growing foundation for the subsequent incorporation of Explainable AI and progressive deep learning systems in neurological health care systems.