Computer Aided Detection System For Renal Calculi Using Effective Segmentation Approaches

Abstract

The main objective of this work is to develop a Computer Aided newlineDetection CAD system for the detection of renal calculi or Kidney Stones newlinefor assisting clinicians in detecting the image of renal calculi and making a newlinetimely decision regarding the treatment procedure newlineIn the case of critical clinical examinations where the number of newlinestones needs to be identified the traditional manual methods become tedious newlineand lack repeatability too The necessity of obtaining high repeatability and newlinethe need for increasing efficiency motivates the development of automated newlineand fast procedures that segment out calculi of different sizes and shapes in newlinemedical images by applying the image segmentation techniques The ultimate newlineaim of medical image segmentation is to reduce the amount of time a newlineradiologist needs to spend for looking at an image to identify the portions of newlinerenal calculi newlineTherefore in the view of addressing aforementioned problem three newlinerenal calculi segmentation methods namely Effective Segmentation of Renal newlinecalculi using Adaptive Neuro Fuzzy Inference System ESRANFIS Inner newlineOuter Regions based new EnhancedWatershed Segmentation method newlineIOREWS and Region Indicator with Contour Segmentation method RICS newlinehave been proposed in this work newlineThe first ESRANFIS segmentation technique for detecting renal newlinecalculi comprises of two main phases namely Preprocessing and newlineClassification phase and Calculi detection phase The first model achieves newlinehigher rate of accuracy in detecting calculi but still other measures need to be newlineimproved for effective detection of renal calculi in a diagnostic process newlineThe second IOREWS method comprising the phases of segmentation newlineof renal calculi using inner outer region indicators improved watershed newlineapproach along with ANFIS is presented with the view of improving other newlinemeasures involved in the detection process The second model achieves newlineconsiderable improvement in all the measures except Specificity and False newlinePositive Rate The third method RICS uses an effective ContourANFIS system for newlinerenal calculi segmentation