A texture analysis approach for kidney boundary detection from ultrasonic images

Boundary extraction from sonograms is an important but difficult issue in medical imaging analysis. Here, the authors propose a new method based on Laws' microtexture energies associated with the MAP (maximum a posteriori) estimation to construct probabilistic deformable model for segmenting kidney boundary from sonograms. It incorporates region-based properties into a contour-based detection mechanism. The optimal contour is obtained by using gradient information subject to a smoothness constraint through the dynamic programming approach. Experimental results show the proposed method successfully detects the renal region.