Evaluation of osteoporosis features in cortical bone models using pattern recognition methods applied to ultrasound data

Abstract

Cortical bone plays a crucial role in bone strength in the context of osteoporosis. Thinning of the cortex and intracortical porosity developing from the endosteum are the main independent indicators of osteoporosis-related fractures risk, which can be separately quantified by advanced imaging techniques. Numerous studies have shown the sensitivity of ultrasonic guided waves to changes in the thickness and porosity of cortical bone both in models and in human subjects. The present study demonstrates feasibility of the pattern recognition approach to separate determination of factors of diagnostic interest – cortical thickness (CTh) and intracortical porosity as the relative thickness of the porous layer (rPTh), where a set of guided waves velocities at different frequencies are used as evaluation criteria. The experiments were carried out on acrylic plastic plates as compact bone phantoms of different thicknesses with simulated pores in the form of CNC-drilled holes from below. The distance-time slopes of fast and slow wave packets at 250, 100 and 50 kHz were obtained as distance-time slopes in the corresponding spatiotemporal waveform profiles built by stepwise acquisition of ultrasonic signals. The evaluation criteria (distance-time slopes) in the training dataset formed the decision rules – 3-dimensional surfaces of the criterion distribution in the field CTh-rPTh. For the test object, the intersection of the segments of the probabilistic solution according to three decision rules gave the evaluation answer. The method allowed us to determine the values of both factors CTh and rPTh with reasonable diagnostic accuracy at different thicknesses of the covering soft tissues. The proposed approach can be adapted for the assessment of osteoporosis after obtaining an appropriate database in human studies.