
AI could transform routine clinical X-rays into a powerful tool to screen for osteoporosis, researchers have discovered. Their findings, in the Journal of Orthopaedic Research, revealed that it was possible to estimate bone mineral density using a system that applied deep learning to X-ray images.
Interestingly, the diagnostic system was able to estimate not only bone mineral density in the lumbar region but also in the femoral region using a plain lumbar anteroposterior X‐ray image. Although it was more accurate for the lumbar region, even its accuracy for the femoral region was deemed “acceptable” by the researchers.
“Bone mineral density measurement is essential for screening and diagnosing osteoporosis, but limited access to diagnostic equipment means that millions of people worldwide may remain undiagnosed,” said corresponding author Toru Moro, PhD, from the University of Tokyo. “This AI system has the potential to transform routine clinical X-rays into a powerful tool for opportunistic screening, enabling earlier, broader, and more efficient detection of osteoporosis.”
At the turn of the century, there were nine million osteoporotic fractures worldwide that included 1.4 million clinical vertebral fractures and 1.6 million hip fractures. The number of hip fractures was projected to increase to 2.6 million by 2020 and to 4.5 million by 2050, with lifetime risks of any hip, spine, or forearm fracture in the United States at 40% in women and 13% in men.
But because osteoporosis is usually asymptomatic, it is common to only receive a diagnosis and start treatment after a fragility fracture.
In an attempt to identify the condition earlier, Moro and team developed an AI‐assisted diagnostic system that calculated lumbar and femoral bone mineral density from anteroposterior lumbar X‐ray images.
They analyzed 1454 X-ray images using the system and reported that the sensitivity of the system—its ability to correctly identify people’s bone density loss—for the lumbar and femur of participants with osteopenia, were 86.4% and 84.1%, respectively.
Its specificity—reflecting the ability to correctly identify those without osteopenia—were a corresponding 80.4% and 76.3%.
The AI-based diagnostic system also showed impressive sensitivity and specificity for categorizing patients with and without osteoporosis.
“Similar sensitivities and specificities to conventional diagnostic systems suggest that this AI‐assisted diagnosis system could become an effective osteoporosis screening tool for the lumbar spine and femur,” the researchers suggested.
They added that it could help prevent bone loss and fragility fractures in areas such as the lumbar spine and femur by encouraging improvements in lifestyle habits, including diet and exercise.





