Abstract

This study presents a Self-Supervised Convolutional Neural Network (CNN) model designed for the early detection and classification of rare neurological disorders using brain MRI images. Early identification of these disorders is critical for effective intervention and improved patient outcomes. Our approach leverages self-supervised learning to enhance the model's ability to extract meaningful features from brain MRI data, even with limited labeled samples. The pipeline includes stages such as data preprocessing, augmentation, model compilation, and training, which help improve the model’s accuracy and generalization. The model’s performance was evaluated using a variety of metrics, including accuracy, precision, recall, and F1-score, showing promising results in distinguishing between "Neurological Disorder" and "No Neurological Disorder" classes. Notably, the model achieved high recall for the disorder class, underscoring its effectiveness in identifying positive cases. This self-supervised CNN framework offers a significant advancement for early classification of neurological conditions in MRI images, especially in scenarios with limited labeled data. Its successful application in this context highlights its potential for clinical integration, providing a valuable tool for neurologists in diagnosing rare disorders early and enhancing patient care.