Implementation of a Model Database for a Novel Ultrasonic Approach to Bone Evaluation

Osteoporosis is considered as a major public health problem, second only to cardiovascular diseases. The gold standard for its diagnosis is currently represented by dual energy X-ray absorptiometry (DXA), which, however, suffers from some important drawbacks. In order to overcome such limitations, the use of ultrasound (US) techniques has been proposed. In this paper, a novel approach to the diagnosis of osteoporosis through US scans on lumbar spine and proximal femur is described. The approach relies on the estimation of diagnostic parameters by measuring the degree of similarity between the spectra of the raw radiofrequency (RF) echo signals and reference spectral models of osteoporotic or healthy bones. Reference models are representative of the features of either osteoporotic or healthy bone structures and are matched with subject age, sex, ethnic group and body mass index to take into account variations in bone physiological condition and subject anatomy. In this paper, the methods implemented to build the database of reference models and to estimate diagnostic parameters are presented. The performance of the approach was assessed on a total of 145 Caucasian underweight and normal-weighted women with age in the range from 46 to 55. Performance was assessed through direct comparison with DXA results. The obtained median relative error in the estimation of bone mineral density was as low as 9.1% on women aged 51 to 55 years and 12.0% on women with age in the range from 46 to 50 years. Moreover, for the two groups, the estimation error was lower than 20% for 81% and 78.6% of subjects, respectively. Therefore, the proposed method combines the advantages of the use of US techniques with a remarkable diagnostic accuracy, thus lending itself to the possibility of being used for population mass screenings.