Automated vertebral compression fracture detection and quantification on opportunistic CT scans: a performance
D Guenoun1, M S Quemeneur2, A Ayobi3
1Department of Radiology, Institute for Locomotion, Sainte-Marguerite Hospital, APHM, 13009 Marseille, France; Institute of Movement Sciences (ISM), CNRS, Aix Marseille University, 13005 Marseille, France.
Clinical Radiology
|February 26, 2025
Summary
A deep learning algorithm accurately screens for vertebral compression fractures (VCFs) on CT scans, aiding in early osteoporosis detection. This tool shows high reliability for opportunistic VCF assessment.
Area of Science:
- Radiology
- Artificial Intelligence
- Medical Imaging
Background:
- Vertebral compression fractures (VCFs) are often asymptomatic and missed on opportunistic CT scans, leading to undiagnosed osteoporosis.
- Early detection of VCFs is crucial for managing osteoporosis and preventing further complications.
Purpose of the Study:
- To develop and evaluate a deep learning (DL) algorithm for opportunistic screening of VCFs on CT scans.
- To assess the algorithm's performance in detecting VCFs, measuring vertebral height loss (VHL), and calculating bone attenuation (mean Hounsfield Units).
Main Methods:
- A DL algorithm using 2D/3D U-Nets was developed to analyze CT scans for VCF detection and quantification.
- The algorithm was trained and tested on retrospective CT scans from 100 patients, comparing its outputs to assessments by two board-certified radiologists.
Main Results:
- The DL algorithm demonstrated high agreement with radiologists for vertebra labeling (94.9%) and VHL measurements (ICC=0.854).
- Excellent correlation was found for mean Hounsfield Units (Pearson's r=0.89).
- The algorithm achieved high sensitivity (92.3%) and specificity (91.7%) for VCF screening.
Conclusions:
- The automated DL tool reliably screens for and quantifies opportunistic VCFs.
- This technology can assist radiologists in VCF assessment and improve opportunistic osteoporosis screening.
