Use of fuzzy techniques for the automatic identification of left ventricle contours from multi-slice short-axis MR imaging

An automatic left ventricle (LV) contouring method in multi-slice short-axis MR imaging making use of fuzzy logic and dynamic programming, is presented. Following the extraction of three parameters for each pixel, the membership degree of each pixel to the cardiac contour fuzzy set is calculated. The first parameter takes into account the pixel grey level, the second the presence of an edge and the third the information collected on the precedent slice. The cardiac contours are detected using dynamic programming on the matrix of membership degrees. The method has been tested on thirty image data sets. This preliminary study clearly demonstrates the interest of this method, especially in order to overcome the principal problems met in the automatic cardiac contour detection on MR imaging, such as poor image contrast or the presence of intra-cavity structures.