Segmentation of tibia bone in ultrasound images using active shape models

This paper describes the use of Active Shape Models to segment the tibia bone in ultrasound images. Using CT images from three volunteers, a training, set consisting of 60 tibia contours, each represented by 45 labeled points, is first established. After aligning the contours in the training set, a Point Distribution Model (PDM) is generated that includes a mean shape of tibia and six modes of shape variation that explain 98% of the total shape variation. Using the PDM, a new shape can be generated which is further scaled, rotated and translated to match the tibia contour in an ultrasound image. Preliminary results indicate that Active Shape Models provide a useful tool for segmenting leg bones in ultrasound images.