Machine Learning for Opportunistic Screening for Osteoporosis and Osteopenia Using Knee CT Scans
Ronnie Sebro1,2,3, Mahmoud Elmahdy2,3
1Department of Orthopedic Surgery, Mayo Clinic, Jacksonville, FL, USA.
Summary
Knee CT scans can identify osteoporosis/osteopenia by analyzing trabecular bone attenuation. Machine learning models using CT data are more effective than single bone measurements for opportunistic screening.
Area of Science:
- Radiology
- Orthopedics
- Medical Imaging
Background:
- Osteoporosis and osteopenia are significant public health concerns.
- Dual-energy X-ray absorptiometry (DXA) is the standard for bone mineral density assessment.
- Alternative, opportunistic screening methods are needed.
Purpose of the Study:
- To predict osteoporosis/osteopenia using trabecular bone attenuation from knee CT scans.
- To evaluate the efficacy of machine learning models for this prediction.
Main Methods:
- Retrospective analysis of 273 patients with contemporaneous knee CT and DXA scans.
- Volumetric segmentation of trabecular bone in knee joints to measure CT attenuation.
- Training and testing of support vector machine (SVM) and random forest (RF) classifiers.
Main Results:
- The proximal tibia showed the best predictive ability for osteoporosis/osteopenia.
- A CT attenuation threshold of 96.0 HU in the proximal tibia yielded an AUC of 0.748.
- The SVM classifier achieved a superior AUC of 0.912, outperforming RF and single-bone thresholds.
Conclusions:
- Knee CT scans enable opportunistic screening for osteoporosis/osteopenia.
- Multivariable machine learning models significantly enhance predictive accuracy compared to single bone CT attenuation.
