Enhancing Kidney Stone Detection Accuracy Using Image Processing and a Hybrid Convolutional Deep Belief Network

Automated CKD is a rapidly expanding non-communicable ailment that significantly contributes to global mortality and morbidity. CKD impacts more than 10% of the worldwide population, leading to millions of deaths annually. This paper proposes an enhanced deep learning model for renal disease detection that integrates convolutional and involution layers to enhance classification accuracy. CKD is a worldwide health issue characterised by the consequences of diminished renal function and renal failure. A KS is a condition that adversely affects renal function. The often asymptomatic nature of this illness necessitates fast and correct identification of renal problems to avert catastrophic consequences. This paper introduces an automated detection system for CT kidney stone images utilising a ConvDBN Model. The suggested study entails the preliminary augmentation of image phases, employing GB + LF, MB + LP, and Bilateral Filter to improve the quality of CT images of the kidneys. GLCM is utilised for feature extraction. Subsequently, ConvDBNs are utilised as a neural network approach for classification, primarily aimed at accurately differentiating between normal and abnormal photographs. Upon an aberrant classification suggesting the existence of a kidney stone, a subsequent measure is enacted. This phase entails the removal of the bed mat and the application of noise reduction techniques to achieve accurate detection and localisation of the kidney stone within the image. The suggested solution attains 98.2% accuracy utilising both quality and noisy image datasets, surpassing previous DL and conventional detection of image methods. To help urologists confirm their physical examination of KS, this automated method can be used to reduce the possibility of human error.