Active Contour driven by geometric mean and standard deviation-based energy fitting model for the left ventricle segmentation from cardiac magnetic resonance images

Abstract

This paper proposes a geometric mean and standard deviation-based energy fitting model to improve the accuracy of segmentation of the left ventricle from cardiac Magnetic Resonance Imaging (MRI). Energy-fitting-based active contour models emerged so far suffer either from intensity inhomogeneity or gives wrong segmentation result due to an inappropriate initial contour. Thus, accurate and robust segmentation of the left ventricle from cardiac MRI still a challenging problem. Therefore, to tackle both the problems, a geometric mean-based energy-fitting model is proposed. Unlike the recent energy-fitting-based models which use the arithmetic mean to calculate the local energy, the proposed method uses geometric mean and scaled standard deviation to compute the energy functional which drives the active contour to the region of interest. In addition to that completely removes the initial contour problem by automating it according to the input. The initial contour in the proposed model is a circle its radius and the center are calculated from the input sample itself. This initial contour is an appropriate and automated one that helps to reduce the computation time for segmentation. Experiments are conducted on cardiac MRI images the result obtained is compared with ground truth and evaluated by Average perpendicular distance (APD) and DICE similarity coefficient. Further the visual, as well as evaluated parameters, evidences that the proposed model performs better than the existing methods.