59598 Automated Quantification of Vertebral Bone Quality Score Using an Artificial Intelligence Framework
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
INTRODUCTION: Automated identification of lumbar spine anatomy on magnetic resonance imaging (MRI) improves clinical workflow and research generalizability. The vertebral bone quality (VBQ) score, based on T1-weighted MRI and normalized to cerebrospinal fluid (CSF), is a surrogate for bone mineral density and fragility risk. METHODS: Lumbar MRI studies were retrospectively collected, screened, and deidentified. Manual VBQ scores were calculated using circular ROIs in L1–L4 vertebral bodies and CSF posterior to L3. Five reviewers performed manual measurements, with each study assessed by two reviewers. Automation used AI-based segmentation to extract vertebral and CSF signal intensities for VBQ calculation. Intraclass correlation coefficients (ICCs) assessed agreement between manual and automated methods. RESULTS: A total of 1,089 patients (mean age 51.5 years, 56% female) underwent MRI across 12 systems; 81% were performed at 1.5 T and 95.6% on GE, Siemens, Philips, or Hitachi machines. Manual and automated VBQ scores were obtained at L1–L4 and CSF. Vertebral body (VB) signal intensities were similar between methods; CSF intensity was higher with automation. Mean reviewer and automated VBQ scores were 3.18 ± 1.20 and 2.98 ± 1.10, respectively. Inter-rater ICCs were 0.930–0.950 for L1–L4 VB, 0.865 for CSF, and 0.789 for VBQ. Reviewer–automation agreement was higher across all measures (VB: 0.981–0.987; CSF: 0.939; VBQ: 0.858). For males ≤50 years, scores were 2.84 ± 0.88 (reviewers) and 2.67 ± 0.89 (automation); for >50 years, 3.24 ± 1.24 and 3.01 ± 1.08. For females ≤50 years, scores were 3.05 ± 1.29 and 2.84 ± 1.19; for >50 years, 3.43 ± 1.21 and 3.26 ± 1.11. CONCLUSIONS: Automated VBQ measurement showed high agreement with manual review and reduced interobserver variability, supporting its use in scalable, reproducible lumbar MRI bone quality assessments.
Fields of Study & Subjects
Medical Imaging and Analysis
Bone health and osteoporosis research
Artificial Intelligence in Healthcare and Education
Publication details
- Journal / source
- Neurosurgery
- Volume
- 72
- Issue
- Supplement_1
- Pages
- 307–307
- Publication date
- 26 March 2026
- Language
- EN