A particle filtering method for muscle tissue segmentation from MRI images

Abstract

We have developed a particle filtering method for muscle tissue reconstruction using the level set representation of contours, which makes it possible to handle topological change of the muscle contour. The inputs of the framework are the selected images which are manually outlined for specified muscle tissue. The contours are calculated using an energy term to predict the likelihood of muscle contour. The priors used in the energy term are acquired using an interpolation method between every two selected images. The results shows that the method can segment muscle tissues with a volume error of 4% and dice coefficient of 0.95 compared with manually outlined results.