Automated deep learning pipeline for measuring lumbar thecal sac AP diameter on mid-sagittal MR images

Key Points

• To develop an automated pipeline for measuring lumbar thecal sac diameter using mid-sagittal MR images.
• Utilized a YOLOv8-OBB model for detection of anatomical structures.
• Incorporated Attention U-Net for enhancing image segmentation.
• Employed mid-sagittal MR images for analysis.
• Achieved approximately 1 mm measurement error compared to expert references.
• Supported reliable and standardized morphometric reporting.
• Enabled assessment of lumbar stenosis through automated measures.