Automated vertebral bone quality score measurement on lumbar MRI using deep learning: Development and validation of
Nishantha M Jayasuriya1, Emily Feng1, Karim Rizwan Nathani2
1Department of Neurologic Surgery, Mayo Clinic, Rochester, MN, USA.
Clinical Neurology and Neurosurgery
|August 8, 2025
Summary
An AI algorithm accurately predicts Vertebral Bone Quality (VBQ) scores from MRI scans, improving preoperative assessment for spine surgery. This technology aids in identifying patients with poor bone health for better surgical outcomes.
Area of Science:
- Radiology
- Artificial Intelligence
- Orthopedics
Background:
- Bone health is crucial for spine surgery success.
- Current bone quality assessment methods are limited.
- Preoperative assessment is vital for optimizing surgical outcomes.
Purpose of the Study:
- To develop and validate an AI algorithm for predicting Vertebral Bone Quality (VBQ) scores.
- To enable automated VBQ scoring from routine lumbar MRI scans.
- To improve preoperative identification of patients at risk for poor surgical outcomes.
Main Methods:
- Utilized 257 lumbar spine MRI scans from the SPIDER challenge dataset.
- Developed a YOLOv8 model for automated region of interest placement and VBQ score calculation.
- Validated the AI system against manual annotations from 47 patients, assessing performance metrics.
Main Results:
- The YOLOv8 model achieved high accuracy in vertebral body detection.
- Strong interrater reliability was observed between human and AI measurements (ICC: 0.88-0.93).
- High Pearson correlations (0.86-0.9) and low RMSE (0.42-0.58) indicate accurate VBQ score prediction.
Conclusions:
- The AI algorithm accurately predicts VBQ scores from lumbar MRIs.
- This approach can enhance early identification of patients with poor bone health.
- External validation is recommended for broader clinical applicability.
