Related Experiment Video
Updated: Apr 11, 2026
07:44
Transforaminal Full-Endoscopic Lumbar Foraminotomy Under Local Anesthesia for L5/S1 Adjacent Segment Foraminal Stenosis
Published on: October 17, 2025
1.8K
Medical Spine Sagittal MRI Dataset for Segmentation and Foraminal Stenosis detection
Osamah F Abdulmahmod1, Mugahed A Al-Antari2, Hyunwook Kwon1
1Department of Artificial Intelligence and Data Science, College of AI Convergence, Daeyang AI Center, Sejong University, Seoul, 05006, Korea.
Scientific Data
|April 9, 2026
Summary
A new dataset of 500 lumbar spine MRIs with expert annotations aids AI development for diagnosing spinal conditions like foraminal stenosis, improving diagnostic accuracy and efficiency.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Spine Surgery
Background:
- Lumbar spine disorders are common, especially in the elderly.
- MRI is crucial for diagnosis but interpretation is complex and time-consuming.
- Developing AI for automated analysis requires high-quality, annotated datasets.
Purpose of the Study:
- To introduce a novel, large-scale sagittal lumbar spine MRI dataset with comprehensive annotations.
- To provide a reliable resource for training and validating AI algorithms for lumbar spine analysis.
- To demonstrate the dataset's utility by developing an AI-based computer-aided diagnosis (CAD) system for foraminal stenosis.
Main Methods:
- Compiled a dataset of 500 sagittal lumbar spine MRIs.
- Included expert-annotated foraminal detection labels (bounding boxes, severity grades) and AI-generated, neurosurgeon-validated segmentation masks for anatomical structures.
- Developed and evaluated an AI-based CAD system for foraminal stenosis assessment.
Main Results:
- The AI-based CAD system achieved 86% accuracy in slice selection, 90% in ROI localization, and 65% in severity classification.
- The SegResNet model achieved high performance in anatomical segmentation (DSC 97.32%, HD95 2.337 mm, recall 97.30%).
- The dataset and annotations were validated for reliability in clinical and research applications.
Conclusions:
- The proposed dataset is a valuable, richly annotated resource for advancing AI in lumbar spine analysis.
- It supports the development and benchmarking of AI algorithms for segmentation, morphological analysis, and automated diagnosis of conditions like foraminal stenosis.
- This resource can significantly aid clinical and research applications in spine diagnostics.
