Développement d’un système de tomographie ultrasonore pour la mesure des propriétés osseuses du radius

Abstract

Bone quality assessment for osteoporosis diagnosis can be performed using a wide array of methods. In clinical applications, the most commonly used technique is dual energy X-ray absorptiometry. However, Li et al. (2013) have shown that this method is inaccurate to diagnose osteoporosis with detection rates averaging 17.1%. The main objective of this project was to investigate the possibility of using an alternate non-irradiating imaging technique to assess bone quality in a view to use the method for the detection of osteoporosis. Ultrasound imaging is a rapidly developing non-irradiating imaging technique. Unfortunately, its use is presently limited to the characterization of materials with low speed of sound contrast relative to the background medium. In this thesis, a new ultrasound imaging method based on the «hybrid algorithm for robust breast ultrasound tomography» proposed by Huthwaite & Simonetti (2011a) is presented. This project focused on adapting the «hybrid algorithm for robust breast ultrasound tomography» for quantitative evaluation of the bone speed of sound. The main objective of this project was to evaluate the possibility of using an adapted «hybrid algorithm for robust breast ultrasound tomography» as a way to quantify speed of sound in bones and therefore assess bone quality. The advantage of this algorithm over traditional tomography is its extended range of applicability. Objects with high velocity contrast relative to the background medium can be imaged using the «hybrid algorithm for robust breast ultrasound tomography». The hypothesis made in this project was that the «hybrid algorithm for robust breast ultrasound tomography» could be adapted to characterize human bone mechanical properties and image bones cross-sections, so as to assess bone quality in a view to develop an osteoporosis diagnosis tool. A 2D finite element model was first developed in ABAQUS to model wave propagation inside a bone phantom immersed in water. A test bench was later used to validate finite element results on bone mimicking phantoms and human cadaveric radius bones. The main difference between the standard «hybrid algorithm for robust breast ultrasound tomography» algorithm and the one developed in this project is the data post-processing algorithm. A new algorithm combining a threshold, the Akaike information criterion and a wave mode separator of the Hilbert transformed time trace was developed to evaluate the time of flight between two transducers with more precision. The images generated using the bone phantom experimental data and the finite element data enabled segregation between osteoporotic and healthy bone phantoms. However, artefacts linked to the density gradient at the water/bone interface in the object function and errors in arrival time estimation were observed. These artefacts reduced imaging quality which decreased image segmentation performances. Incorporating a density map or machine learning algorithm could increase image precision by reducing errors linked with rapid density change and time of flight estimation.

Date	8 Feb 2018
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Pierre Bélanger (Supervisor)