Paper Title
Artificial Intelligence Segmentation Framework for Identifying Significant Pathological Areas Causing Lumbar Spinal Stenosis

Abstract
Lumbar Spinal Stenosis (LSS) is a condition characterized by the narrowing of the spinal canal in the lumbar region, resulting in pressure on the spinal cord and nerves. This can cause symptoms like pain, numbness, weakness, or tingling sensations in the lower back, buttocks, and legs. In this study, we introduce a new AI-based multi-class segmentation framework designed to identify significant pathological regions, including the Intervertebral Disc (IVD), Posterior Element (PE), Thecal Sac (TS), and the Area between Anterior and Posterior (AAP) vertebrae. The proposed computer-aided diagnosis (CAD) framework comprises four main steps: medical MRI data acquisition and collection, preprocessing, optimization, selection of a backbone XAI model, and quantitative and qualitative performance evaluation. We utilized the Mendeley public benchmark LSS dataset to build and train the proposed AI segmentation framework. Meanwhile, the entire pipeline is verified and validated using our private MRI dataset (i.e., AISSLab LSS MRI dataset) where it is collected under the international research collaboration protocol between Korea and Turky. Among the three segmentation AI models tested (UNet, SwinUNETR, and UNet++), the UNet model demonstrated slightly better segmentation performance, achieving a class-wise dice scores of97.51% (±0.014), 92.78% (±0.033), 90.75% (±0.0718), and 78.88% (±0.0734) in terms of IVD, PE, TS and APP, respectively. Our proposed framework exhibits promising reliability and trustworthiness for medical industrial applications in LSS segmentation. Keywords - Lumber Spinal Stenosis (LSS), Back Pain, Herniated Spinal Disk, Nerve Compressing, AI-based Segmentation, Spinal MRI Axial Images, Explainable AI (XAI).