Abstract
<jats:p>Recent encroachments in Artificial Intelligence (AI) devise potential in automating the classification of lumbar spine degeneration, especially using deep learning models trained on Computed Tomography (CT) and Magnetic Resonance Imaging (MRI) scans. The most common causes of lower back pain are Disc herniation, spinal stenosis, and spondylolisthesis in world wide. So for efficient treatment planning it is important to have the right diagnosis. However, manually interpreting imaging data is still time-consuming, subjective and might vary from one radiologist to the next. But only AI models are black boxes, they can't be used in clinical settings because they aren't clear or easy to understand. To tackle this issue, Explainable AI (XAI) methods like SHAP and LIME have been proposed to spine classification frameworks to visual and feature-based reasons for model predictions. This study shows how XAI improves clinical trust by locating important imaging areas, measuring the contributions of radiological features and ensuring AI results accuracy with clinical reasoning to managing lumbar spine degeneration. in line with medical reasoning.</jats:p>