Abstract
Muscle architecture features, typically defined by the length, curvature and the angle of insertion of fascicles, are directly related to muscle function. In addition to its applications in fundamental and applied physiology, these features are relevant in several clinical contexts such as aging, obesity, cerebral palsy or critical illnesses in general. These features can be extracted from ultrasound images. Manual analysis of the ultrasound images is, in addition to being time-consuming, associated with a high degree of uncertainty when it comes to reliability. An automatic algorithm for detection of these features would not only be time-saving, but also ensure objectivity which will contribute to the reliability of the estimations. This thesis proposes an algorithm for estimating pennation angle and fascicle length in single frame ultrasound images of the vastus lateralis. This algorithm includes detection of the region of interest, filtering of the images to reduce speckle noise and enhance the structures, aponeuroses detection, fascicle detection and calculation of the muscle features. The main aim of filtering the image is to enhance the structures in the image while reducing the noise. We best method to achieve this is using the Knutsson tensor filters. The detection of aponeuroses is done using local Radon transform, while the detection of fascicles is done using a normalized local Radon transform. The pennation angle is estimated based on the angle of the lower aponeurosis and a representative reference fascicle, resulting in one pennation angle. The fascicle length is estimated based on a reconstructed fascicle represented by a 2. degree polynomial, taking the curvature into account. The aponeuroses are represented as straight lines. The aponeuroses and fascicles are extrapolated in cases where the entire fascicle is not within the field of view. The algorithm shows promising results, but further improvements should be made in order to make it more reliable, especially when it comes to how the aponeuroses and fascicles are represented.