Automatic landmark detection on Trochanter Minor in x-ray images

Abstract: During pre-operative planning for hip replacement, the choice of prosthesis can be aided by measurements in x-ray images of the hip. Some measurements can be done automatically but this require robust and precise image processing algorithms which can detect anatomical features. The Trochanter minor is an important landmark on the femoral shaft. In this thesis, three di.erent image processing algorithms are explained and tested for automatic landmark detection on Trochanter minor. The algorithms handled are Active Shape Models, Shortest path algorithm and a segmentation technique based on cumulated cost maps. The results indicate that cumulated cost maps are an e.ective tool for rough segmentation of the Trochanter Minor. A snake algorithm was then applied which could .nd the edge of the Trochanter minor in all images used in the test. The edge can be used to locate a curvature extremum which can be used as a landmark point.