Opportunistic Screening for Osteoporosis Using Hand Radiographs - A Feature Augmentation Study Technique (FAST)

Area of Science:

• Radiology
• Osteoporosis Research
• Artificial Intelligence in Medicine

Background:

• Low bone mineral density (BMD) increases fracture risk but is often undiagnosed.
• Opportunistic screening for low BMD during other medical evaluations is crucial.

Purpose of the Study:

• To predict osteoporosis/osteopenia using hand radiographs.
• To correlate bone textural analysis from hand X-rays with dual-energy X-ray absorptiometry (DXA) measurements.
• To assess the viability of hand radiographs as a screening tool for low BMD.

Main Methods:

• Retrospective study of 422 patients aged ≥ 50.
• Utilized hand radiographs and DXA scans from three vendors.
• Employed a deep learning model trained on a dataset divided into training/validation and test sets.

Main Results:

• The deep learning model achieved 77.38% accuracy, 77.38% sensitivity, and 73.63% specificity.
• The model demonstrated an Area Under the Curve (AUC) of 83% for detecting osteoporosis/osteopenia.
• Significant correlations were found between bone textural analysis and DXA measurements.

Conclusions:

• Hand radiographs, analyzed with deep learning, show potential as a screening tool for osteoporosis/osteopenia.
• This approach can identify individuals who may benefit from formal DXA assessment.
• Opportunistic screening using hand X-rays could improve early diagnosis and management of low BMD.