Automated tracking of muscle fascicle orientation in B-mode ultrasound images

B-mode ultrasound can be used to non-invasively image muscle fascicles during both static and dynamic contractions. Digitizing these muscle fascicles can be a timely and subjective process, and usually studies have used the images to determine the linear fascicle lengths. However, fascicle orientations can vary along each fascicle (curvature) and between fascicles. The purpose of this study was to develop and test two methods for automatically tracking fascicle orientation. Images were initially filtered using a multiscale vessel enhancement (a technique used to enhance tube-like structures), and then fascicle orientations quantified using either the Radon transform or wavelet analysis. Tests on synthetic images showed that these methods could identify fascicular orientation with errors of less than 0.06°. Manual digitization of muscle fascicles during a dynamic contraction resulted in a standard deviation of angle estimates of 1.41° across ten researchers. The Radon transform predicted fascicle orientations that were not significantly different from the manually digitized values, whilst the wavelet analysis resulted in angles that were 1.35° less, and reasons for these differences are discussed. The Radon transform can be used to identify the dominant fascicular orientation within an image, and thus used to estimate muscle fascicle lengths. The wavelet analysis additionally provides information on the local fascicle orientations and can be used to quantify fascicle curvatures and regional differences with fascicle orientation across an image.