Using time information to make a robust feature extraction and parameter estimation from unipennate muscles in B-mode ultrasound sequences.

Abstract

Ultrasound imaging for medical applications is considered a cost-effective and non-invasive method of studying the internal structures of the body. One application is to study muscles and muscle architecture. Muscle architecture refers to a muscle and the arrangement of structures essential to the muscle’s function. One aspect of studying muscle architecture is to study how fascicles move and the fascicles length and pennation angle relative to an aponeurosis. These measurements are used to give us information on force production. In this thesis, the aim was to create an algorithm for B-mode ultrasound sequences that could detect the shallow and the deep aponeurosis and create a robust reference fascicle that could be used to measure the changes in pennation angle and fascicle length over time. We did not manage to create a robust method for detecting the deep aponeurosis, but we were able to find and detect the shallow aponeurosis in each frame consistently. We also developed a method of using à priori knowledge to find the location of the aponeuroses faster than previously. The construction of the reference fascicle as proposed in this thesis seems to be a robust way of creating a reference fascicle especially in sequences were some of the frames have little relevant information.