Optimal Geodesic Active Contours: Application to Heart Segmentation

We develop a semiautomated segmentation method to assist in the analysis of functional pathologies of the left ventricle of the heart. The segmentation is performed using an optimal geodesic active contour with minimal structural knowledge to choose the most likely surfaces of the myocardium. The use of an optimal segmentation algorithm avoids the problems of contour leakage and false minima associated with variational active contour methods. The resulting surfaces may be analysed to obtain quantitative measures of the heart’s function. We have applied the proposed segmentation method to multislice MRI data. The results demonstrate the reliability and efficiency of this scheme as well as its robustness to noise and background clutter.