The goal of this study is to assess the performance of an artificial intelligence software (Osteo Signal) in detecting osteoporosis risk in adults 50 years and older. The main question it aims to answer is: What is the accuracy of the software in detecting osteoporosis risk on chest x-ray images as compared to the standard technique of dual-energy x-ray absorptiometry (DXA)? There is no direct involvement of participants in this study as it will use data from individuals who have already had a chest x-ray and a DXA scan taken in the past.
Age ≥50 years at the time of chest X-ray
Availability of
Posteroanterior (PA) Chest X-ray and
DXA scan with a valid T-score measurement where both the DXA scan and Chest x-ray examinations were performed within 6 months (182 days) of each other and
DXA scan inadequacy: Missing T-score from a technically inadequate scan.
PA Chest X-ray technical inadequacy: Non-PA projection, significant motion artefacts, inadequate visualisation of thoracic structures, or missing DICOM metadata.
Image format incompatibility: Non-DICOM format or corrupted image files that would prevent AI analysis
PA Chest X-ray > ±6 months from the date of the DXA scan.
Patients selected hospitals in the Netherlands and UK that have previously had chest x-rays and DXA scans
Cohort
Retrospective
Osteo Signal
DeviceArtificial Intelligence (AI) software
Two class bone status classification (osteoporosis/non osteoporosis)
Device output against the lowest DXA as measured by sensitivity, specificity, and area under the ROC curve (AUC)
Three class bone status classification (osteoporosis/osteopenia/normal)
Device output against the lowest DXA as measured by area under the ROC curve (AUC), overall per cent agreement, and weighted Cohen's kappa
Classification for osteoporosis and osteopenia
Positive predictive value (PPV) and negative predictive value (NPV) for osteoporosis and osteopenia classifications
Click on the lead sponsor to view all of their trials.
