An Automatic Approach for The Detection and Segmentation of Kidney Stone In Kub CT Images Using Mask R-CNN

Abstract

Kidney stones, medically known as nephrolithiasis, are a common urological condition affecting individuals worldwide. In uriolithiasis, unwanted sediment deposits in the kidneys, affecting the normal function of the kidneys and urination system. Urine blockage and urinary tract infection (UTI) are very common in patients with urolithiasis. In order to provide accurate diagnoses and effective treatment plans for kidney stones, it is crucial to develop proper medical imaging techniques. In recent years, deep learning algorithms have been developed for the accurate detection and segmentation of stones in medical images. Prior to this work, no such system had been developed to detect and segment kidney stones in CT KUB images using Mask R-CNN deep learning algorithm. This study aims to present an automated method for the accurate detection and segmentation of kidney stones in CT KUB (Kidneys, Ureters, and Bladder) images. In this study, the state-of-the-art deep learning model Mask R-CNN (Masked Region-based Convolution Neural Network) has been used to detect and segment the kidney stones in KUB CT images. In order to segment the stones effectively, both ResNet50 and ResNet101-based Mask R-CNN architectures were deployed. We used a dataset consisting of 15000 KUB CT images to develop a model for detecting and segmenting kidney stones. The performance of proposed model was evaluated using Accuracy, Precision, Recall, and F1 Score. Experimental results shows 96.9% precision, 88.0% recall and 92.0% F1 score. Overall 91% mean average precision (mAP) was achieved for the kidney stone detection.