MRI Lumbar Spine Disease Classification Using Deep Visual Transformer Model

Low back pain, as noted by the World Health Organization, is the leading cause of disability worldwide, affecting 619 million people in 2020. This condition affects most people at some point in their lives, with its prevalence increasing with age. Magnetic resonance imaging (MRI) is currently used to provide detailed views of the lumbar spine, enabling radiologists to assess the presence and severity of various conditions. This research aims to detect and classify degenerative lumbar spine conditions using deep learning and ViT model, which may assist radiologists in making accurate diagnoses, determining appropriate treatments, and considering potential surgeries to alleviate back pain and improve patients' overall health and quality of life. The study explored the use of the Visual Transformer (ViT) model for detection and classification. It focused on a five-class lumbar disease classification problem using MRI images. The evaluation compared to traditional machine learning algorithms, revealing superior performance in terms of accuracy, precision,f1_score and recall value. Our model achieved an impressive 97% accuracy on the testing set and 99% on the training set, indicating its potential for accurate spinal lumbar disease classification.