JMIR Publications

ABSTRACT

Background:

Osteoporosis (OP) is poised to become a major issue significantly impacting the well-being of middle-aged and old individuals. Machine learning (ML) and deep learning (DL) models developed based on medical imaging have enhanced clinicians’ diagnostic accuracy and work efficiency. However, the diagnostic performance of different types of medical imaging for OP remains elusive.

Objective:

By summarizing the literature to elucidate the role of DL models based on different medical imaging modalities in detecting OP.

Methods:

PubMed, Embase, Cochrane Library, and Web of Science were retrieved on May 16, 2024, utilizing subject headings and free-text terms, without restrictions on region and study type. A bivariate mixed-effects model was employed for the meta-analysis of sensitivity (SEN) and specificity (SPC). Subgroup analyses were conducted based on ML type, validation set generation method, and examination parts.

Results:

Our systematic review and meta-analysis encompassed 60 studies with 66,195 cases. Among these, 37 studies utilized computed tomography (CT)-based image learning, 22 studies used X-ray-based image learning, and 3 studies employed magnetic resonance imaging(MRI)-based image learning. For diagnosing OP, the SEN and SPC of X-ray-based ML models were 0.92 (95% CI: 0.88 - 0.94) and 0.83 (95% CI: 0.76 - 0.88). The SEN and SPC of CT-based ML models were 0.91（95% CI：0.89-0.93）and 0.92（95% CI：0.89-0.94）. Three studies have constructed diagnostic models for OP based on MRI. The SEN of these models was 0.857, 0.872, and 0.892, and the SPC was 0.944, 0.688, and 0.892, respectively.

Conclusions:

Image-based ML, particularly DL models based on X-ray and CT, demonstrates good accuracy in diagnosing osteoporosis.