- Clinical validation study including 25,000 scans from five multinational cohorts
- Study presentation to take place at ASBMR 2024 congress
- Lead investigator Guillaume Gatineau to receive an award from the ASBMR, recognizing his significant contributions to the field
Geneva, Switzerland — September 27, 2024
Medimaps — the creator of TBS® (Trabecular Bone Score®), the global standard in bone microarchitecture assessment — today announced the presentation of new data on its AI deep learning software model to opportunistically detect bone fragility from conventional X-ray imaging, at the American Society for Bone and Mineral Research annual congress (ASBMR) in Toronto, Canada, September 27–30.
"We know that 70%1 of patients at high risk of fragility fracture have never been referred for a DXA scan to address their osteoporosis. By developing an opportunistic screening tool, our goal is to address this gap. Our new software model uses patients' routine X-ray scans, taken for other pathologies, to analyze their bone microarchitecture and density — requiring no further examination time for patients or radiologists. It flags patients who might present high bone fragility risk, enabling their referral for further DXA examination. This is a smart way to use AI for early detection and subsequent prevention of osteoporotic fractures."
said Prof. Didier Hans, co-founder and CEO of Medimaps.
The new data will be presented by lead investigator Guillaume Gatineau, PhD candidate in AI at the Interdisciplinary Center of Bone Diseases at Lausanne University Hospital, Switzerland, and Data Scientist at Medimaps Group, across the following sessions:
- Plenary Poster Session I — September 27, 5:30–7:30 PM
- Early Career Spotlight Rapid Fire Presentation — September 28, 1:45–2:15 PM
- Poster Session 1 — September 28, 2:15–3:45 PM
Focused on minimizing false positives, the AI model underwent rigorous training and validation on over 25,000 lumbar-spine X-ray DICOM files. It demonstrated an internal validation accuracy of 86% and an external validation accuracy of 85%. These results reflect a significant reduction in false positives, an important factor in preventing unnecessary referrals and addressing a common limitation of over-diagnosis seen in some other AI diagnostic tools.
For this study, Guillaume Gatineau will receive a Young Investigator Award from the ASBMR scientific committee in Toronto.
"Our opportunistic software's robust specificity underscores its capacity to reduce false-positive rates, an important criterion for hospital centers and private practices when purchasing new AI applications. We expect to present further validation studies confirming the clinical relevance, generalizability, and efficacy of our multi-modal software under development in the near future."
added Prof. Didier Hans.
Globally, over 37 million fragility fractures occur annually due to osteoporosis. The challenge lies in the under-identification of individuals at risk, resulting in missed opportunities for preventive treatment.
References
1. Gatineau, G., De Gruttola, M., Nguyen, G., Davies, M., Hind, K., Kužma, M., Payer, J., Guglielmi, G., Fahrleitner-Pammer, A., Krueger, D., Binkley, N. and Hans, D. *Artificial Intelligence-Based, Automatic and Explainable Bone Fragility Detection from X-ray: a Multi-National Cohort Study.* Oral presentation, ASBMR Annual Congress 2024.