Dual X-ray absorptiometry (DXA) and the FRAX score lack sensitivity to identify postmenopausal women at high risk of fracture. To overcome this shortcoming, evaluation of bone microarchitecture using high-resolution peripheral quantitative computed tomography (HR-pQCT) has been suggested to improve fracture risk prediction.
In several prospective studies, bone microarchitectural parameters, evaluated using the finite element analysis (FEA) method, have provided better prediction of fracture risk than BMD alone, measured using DXA, or FRAX score. Most cohorts with prospective data have been combined in the Bone Microarchitecture International Consortium analysis, which confirmed on a large scale the improvement of fracture risk prediction, especially with FEA at the radius, but the magnitude of the improvement was not substantial. A recent study has shown that analyzing the microarchitecture to identify women to treat was cost-effective when using zoledronate. A deep learning model using only the images of the distal forearm, including both the bone and soft tissues, has also improved fracture risk prediction substantially.
The adoption of deep learning to analyze bone microarchitecture is likely to simplify and speed up the process of fracture risk evaluation. This will allow for adequate preventive therapy of a large proportion of postmenopausal women at high risk who are currently left untreated.