Kidney Stone Detection using Amalgam of ML

The prevalence of kidney stones, or urolithiasis, has surged globally over the years, impacting individuals' quality of life due to the potential for painful urination if left untreated. Leveraging the intrinsic advantages of machine learning (ML), such as automated learning, handling multidimensional data, and continuous refinement, has gained significant attention in addressing this issue. Previous research efforts have employed various ML techniques to enhance the prediction of kidney stone occurrence. However, there remains a need to augment accuracy in prediction models. To address this gap, we propose the utilization of Enhanced Modified Binary Particle Swarm Optimization (MBPSO) for feature selection, with the goal of improving accuracy. In the proposed system, Enhanced Modified XGBoost (eXtreme Gradient Boosting) is employed to predict the severity of kidney stone cases, incorporating customized loss functions to enhance interpretability and refine model accuracy. Internal evaluations demonstrate superior predictive performance of the proposed system compared to Decision Tree (DT) and Naïve Bayes (NB) methods in forecasting kidney stone severity.