The renal vessel segmentation for facilitation of partial nephrectomy

Abstract

In this article we have proposed several image processing techniques enabling the extraction of 3D tumor affected renal vascularity from CT scans in order to facilitate partial nephrectomy. The information which vessels supply the tumor is crucial to eliminate ischemic injury and allows the usage of the selective clamping method. However, until now renal vascularity has been analyzed only on the basis of visualization and its limitations. Our novel method consisted of the following steps: binarization upon image intensity histogram, erosion - elimination of connections between different structures, segmentation by a proposed locally adaptive region growing algorithm and finally segmentation by level set method using variational approach allowing the incorporation of the Chan - Vese model and image gradient information into the energy functional. The proposed set of image processing techniques allowed us to obtain 3D renal vessels segmentations and to identify target vessels. The results were validated on manually segmented, randomly chosen slices of ten different patients' computed tomography scans. Segmentation effectiveness is equal to 0.838 of Dice Coefficient meaning.