A knowledge guided left ventricular segmentation-the Taguchi approach

The deformable model (snake) is widely applied in segmentation for medical images especially for cardiac image analysis. Although the model is flexible in the contour deformation process, the weights of associated energy in the model are always given empirically. The systematic (or automatic) weight assignment is still unavailable. In this paper, the authors propose a new method to determine the weights of snake properly for the detection of left ventricle from sonogram. The Taguchi method is adopted to obtain the weight assignment efficiently. To avoid the type one error in weight assignment, the ANOVA test is performed to verify the correctness of the authors´ assumption. The experimental results show that the proposed method successfully resolves the weight assignment problem in the deformable model and obtains satisfactory cardiac contour detection