AI analysis of chest X-rays to screen for osteoporosis in adults 50 and older

A Retrospective Multi-centre Paired Accuracy Study to Evaluate the Performance of Osteo Signal Against the Dual-energy X-ray Absorptiometry (DXA) for Opportunistic Osteoporosis Screening.

This project will test whether an AI tool can use routine chest X-rays to find people aged 50+ who are at risk for osteoporosis compared with standard DXA scans.

Quick facts

• Study type: Observational
• Enrollment: 595 (estimated)
• Ages: 50 Years and up
• Sex: All
• Sponsor: Promedius Inc (Industry-sponsored)
• Locations: 3 sites (Rotterdam and 2 other locations)
• Trial ID: NCT07421713 on ClinicalTrials.gov

What this trial studies

This observational, retrospective project applies the Osteo Signal AI software to existing PA chest X-rays and compares the AI output to DXA T-scores obtained within six months of the X-ray. Eligible records include adults aged 50 or older with both a valid chest radiograph and a DXA scan performed on or after 2010, plus complete demographic metadata. Poor-quality or non-PA X-rays, incompatible image files, and patients already diagnosed or treated for osteoporosis before the DXA are excluded. Data are drawn from participating hospitals in the Netherlands and the UK, and there is no active intervention or new imaging for participants.

Who should consider this trial

Good fit: Adults aged 50 or older who have a PA chest X-ray and a DXA scan with a valid T-score performed within six months of each other, with complete age and sex metadata and no prior osteoporosis diagnosis or treatment before the DXA.

Not a fit: People without paired chest X-ray and DXA data at a participating center, with non-PA or poor-quality images, under age 50, or already treated for osteoporosis before the DXA are unlikely to benefit from inclusion.

Why it matters

Potential benefit: If successful, the approach could help detect undiagnosed osteoporosis earlier by using chest X-rays that patients already had, potentially prompting timely bone-density testing and treatment.

How similar studies have performed: Prior research using AI for opportunistic bone-density screening on imaging (including CT and X-rays) has shown promising but variable accuracy, so this approach has precedent but is not yet standard practice.

Eligibility criteria

Inclusion Criteria:

1. Age ≥50 years at the time of chest X-ray
2. Availability of

   1. Posteroanterior (PA) Chest X-ray and
   2. 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
   3. Available X-ray and DXA scans performed on or after 01-Jan-2010.
3. Complete metadata (sex and age) available on PACS or hospital records.

Exclusion Criteria:

1. Non-diagnostic Quality or Incomplete Imaging of:

   1. DXA scan inadequacy: Missing T-score from a technically inadequate scan.
   2. PA Chest X-ray technical inadequacy: Non-PA projection, significant motion artefacts, inadequate visualisation of thoracic structures, or missing DICOM metadata.
   3. Image format incompatibility: Non-DICOM format or corrupted image files that would prevent AI analysis
   4. PA Chest X-ray \> ±6 months from the date of the DXA scan.
2. Previous diagnosis or treatment of Osteoporosis prior to the DXA scan used in the Study.

   1. diagnosis of osteoporosis (T-score ≤-2.5), or current/recent anti- osteoporotic medications, within 12 months prior to the DXA scan used in the study.
   2. Current or recent anti-osteoporotic medications for 3 months (91 days) continuously prior to the Chest x-ray used in the Study.
3. Diagnosis of any metabolic bone disease prior to the DXA scan used in the study.
4. Diagnosis of metabolic bone disease other than Osteoporosis or Osteopenia, up to a maximum of 3 months (91 days) after the DXA scan.
5. Major Chest Wall or Spinal Deformities, including but not limited to:

   1. Severe structural deformities: Thoracic scoliosis with Cobb angle \>40°, severe kyphosis, or chest wall deformities (pectus excavatum/carinatum) affecting \>30% of the thoracic cage.
   2. Surgical hardware: Previous thoracic/spinal surgery with metallic implants obscuring anatomical landmarks required for AI analysis.
   3. Pathological conditions: Extensive pulmonary consolidation, pleural effusion, lesions or masses obscuring \>30% of thoracic skeletal structures
6. Subject has opted out of confidential data being used for research purposes

Where this trial is running

Rotterdam and 2 other locations

• Erasmus University Medical Center Rotterdam, — Rotterdam, Netherlands (Not_yet_recruiting)
• Mid and South Essex NHS Foundation Trust — Westcliff-on-Sea, Essex, United Kingdom (Not_yet_recruiting)
• Barts Health NHS Trust — London, United Kingdom (Recruiting)

Study contacts

• Principal investigator: Jacob J Visser, Musculoskeletal Radiologist — Erasmus Medical Center
• Study coordinator: Kim Bensalem

Conditions Osteoporosis, Osteopenia; Osteoporosis; Bone Diseases, Metabolic