Characterization of Renal Parenchymal Lesions on Computed Tomography Images Using Texture Analysis

Abstract

This study aim is to characterize renal parenchymal selected lesions (renal cysts and renal cell carcinoma) from normal renal parenchyma using texture analysis this study chosed on basis of a current knowledge gab. The study is an analytical case-control retrospective design it’s population is computed tomography images for normal renal parenchyma, renal cysts and renal cell carcinoma (sample choice was recommended by ministry of health ethical approval committee); The sampling technique is convenient random withdrawn from three institutes (Modern medical centre, Antalya medical centre and Ibn Sinaa specialized hospital) the sample size is 288 images, studied using first order texture features and then studied using higher order texture features the images inclusion criteria is being axial non-enhanced renal parenchyma images. The study excluded any images for the same individuals which did not contain any of the region of interest parts or not deferentially diagnosed to be normal, affected with renal cyst or renal cell carcinoma. The data collected within the time frame starting from December 2018- December 2020. The results of this study is high discrimination power of the textural features (variance, energy, entropy) from the first order texture features group which is (96.5%) accurate in defining the study sample and also high discrimination power of the textural features: Long runs emphasis, Gray Level Non-uniformity, Run Length Non-uniformity, Run Percentage, low Gray Level Run Emphasis, High Gray Level Run Emphasis, Short Run Low Gray Level Emphasis, Long Run Low Gray Level Emphasis, from the higher order texture features group, The accuracy of those features is (93.7%). The study concluded two accurate models with (96.5%), (93.7%) accuracy based on first order feature and higher order features respectively for describing the predefined groups. The main limitation of the study is availability of data that matches the inclusion criteria; hence the researcher suggests future studies with larger data sets.