Suitability of low-frequency axial transmission acoustics as a screening method for bone mass density-defined osteoporosis
Taylor, William R.
Kramers-de Quervain, Inès A.
- Other Conference Item
Purpose: Osteoporosis is a ubiquitous challenge for society, health care providers and the economy in general, affecting 200 million women worldwide . The gold-standard for diagnosing osteoporosis is dual x-ray absorptiometry (DXA), which due the involved radiation exposure is unsuited for routine check-ups. This, together with DXA’s limited availability caused by expensive equipment, low portability, and the need for highly trained personnel, leads to under-detection of osteoporosis, especially in non-developed countries and for people younger than 65 years [2,3]. Quantitative acoustic (QA) methods have recently received renewed attention as a promising non-radiative, non-invasive, inexpensive, and portable assessment technique of bone health. While most research focusses on using QA to gather bone information inaccessible by DXA, the presented work investigates the suitability of axial transmission QA as a screening approach for DXA-defined osteoporosis, thereby improving resource allocation in the clinics as well as bringing osteoporosis early detection to a general health practitioner level and to non-developed countries. Materials and Methods: 40 female subjects above the age of 65 years were measured using axial quantitative acoustics (ax-QA) at the tibia and using dual energy x-ray absorptiometry (DXA) at the hip, distal radius, and spine. The accuracy of classifying the DXA-based osteoporotic state was quantified for the following ax-QA classifiers: a) a simple threshold classifier based on the extracted phase velocity values and b) a support vector machine (SVM) classifier based on the raw acoustic signals. Results and Discussion: The Receiver-Operating Characteristic for the threshold classifier resulted in an area-under-curve value of 0.78 (see Fig. 1). The median and interquartile range of the SVM’s classification accuracy determined by subject-grouped 8-fold cross-validation was 0.63 and 0.19. While the SVM’s performance lags behind expectations, indicating that further development of this approach is required, the performance of the threshold classifier appears suitable for widespread screening application Show more
External linksSearch via SFX
Organisational unit03994 - Taylor, William R.
MoreShow all metadata