Hip fractures are a significant event for elderly patients, often leading to decreased functionality and increased morbidity within one year.
But osteoporosis is often underdiagnosed and undertreated because of the cost and complexity of dual-energy X-ray absorptiometry (DEXA) scans used to gauge the risk for osteoporosis. However, AI and deep learning (DL) models can help predict hip bone mineral density in patients without a complicated scan, according to a study presented at RSNA 2020.
DEXA screenings can diagnose osteoporosis before a bone is broken or hip is fractured by measuring bone mineral density. But declining reimbursement for DEXA screenings and the difficulty of care coordination between orthopedic and primary care physicians may deter patients from getting the exam.
In his presentation, Justin Krogue, MD, a radiologist in the Department of Orthopaedic Surgery at the University of California San Francisco, explained how AI can help solve the problem.
"This study investigates the performance of a deep learning-based algorithm trained to predict bone mineral density directly from plain hip X-rays," Dr. Krogue said.
The study looked at 3,479 hip or pelvic X-rays from 1,337 patients who also had a DEXA scan. Eighty percent of the patients were used to train the model and 20% were used to test it. Total hip bone mineral density (BMD) values were taken from the DEXA scans to train the model to predict the BMD from the plain X-ray.
Bone density results are reported using T-scores, which compares the patient's bone density to that of a healthy 30-year-old adult. According to the World Health Organization, a T-score of -1 or above is normal, -1 to -2.5 is osteopenia, -2.5 or below is osteoporosis.
In the study, 16% of patients were osteoporotic and 43% were osteopenic according to their T-scores. The DL model predicted bone mineral density within 0.1 of the actual value in 74.4 percent of cases, and within 0.15 in 88.0 percent of cases.
"We demonstrated that deep learning is able to identify osteoporosis and osteopenia from X-rays with high levels of accuracy," Dr. Krogue said. "This has the potential to improve patient outcomes by increasing detection and treatment of patients with osteoporosis."
For More Information:
View the RSNA 2020 session Prediction of Hip Bone Mineral Density from Plain Radiographs with Deep Learning — SSMK06 at RSNA2020.RSNA.org.
