SEMI-AUTOMATED KIDNEY MRI SEGMENTATION USING OPTIMAL THRESHOLDING TECHNIQUE (OTT) IN ADAPTIVE NLM FILTERING

Abstract

Biomedical image processing is an important field of medical image in analyzing, and determining the diameter, volume, and vasculature of a tumor or organ. Kidneys are bean-shaped organs that filter blood and excrete waste, maintaining blood pressure as well as eliminating toxins. Segmentation Accuracy plays a vital role in quantifying kidney structure and volume to estimate the kidney disease and also preprocessing is an important stage in segmentation process. Moreover, Clinical methods are not useful to find GFR for single Kidney when compared to imaging techniques. To improve the segmentation accuracy for Kidney MRI images, to quantify the Surface area and Volume of Kidney Compartments to estimate the kidney disease, an adaptive NLM algorithm with new proposed thresholding is adapted to denoising the MRI image and a fast and simple Chan-Vese (CV) levelset formulation used to split the kidney structure from its background. Further, a hard clustering technique used to segment the kidney compartments. The overall performance evaluated in performance metrics such as segmentation Accuracy, Sensitivity, MCC, Dice Coefficient and Jaccard Index. The performance metrics compared with Threshold based FCM. The segmentation results were shown best accuracy nearly 99.7% and very good performance metrics.