AI-Driven Urine Analysis for Non-Invasive Kidney Stone Detection: A Machine Learning Approach

The urological problem of kidney stones becomes a serious health concern when left untreated that may lead to unbearable pain and important medical complications. Patients often face two issues when using traditional diagnosis methods such as CT scans and ultrasounds: these tests cost a large amount of money and expose people to radiation while also being unavailable in areas that lack resources. The research evaluates a machine learning solution which examines urine test results to identify kidney stones because it aims to develop a cost-effective diagnostic method that operates through non-invasive methods with open accessibility. Among the machine learning classifiers used in this study, Logistic Regression, Support Vector Machine (SVM), Random Forest and XGBoost were included because they are known to do well in biomedical classification. Analysis of the urine was done using six important factors: specific gravity, pH, osmolality, conductivity, urea and calcium. Random Forest had the best results, with an accuracy of 98.0 %, precision of 0.989, recall of 0.968, F 1 -score of 0.978 and an AUC-ROC score of $\mathbf{0. 9 8 6}$. XGBoost gave satisfactory results, but Random Forest did better at balancing sensitivity and specificity, making it the best option for early detection in the clinic. The results strengthen the idea that AI might help with kidney stone diagnosis without the need for imaging, especially when money is tight or proper care is far away. By using machine learning with biochemical tests in a thoughtful manner, it is possible to improve detection earlier and to reduce the number of invasive procedures.