Development and external validation of a machine learning-based model to classify uric acid stones in patients with kidney stones of Hounsfield units < 800.

The correct diagnosis of uric acid (UA) stones has important clinical implications since patients with a high risk of perioperative morbidity may be spared surgical intervention and be offered alkalization therapy. We developed and validated a machine learning (ML)-based model to identify stones on computed tomography (CT) images and simultaneously classify UA stones from non-UA stones. An international, multicenter study was performed on 202 patients who received percutaneous nephrolithotomy for kidney stones with HU < 800. Data from 156 (77.2%) patients were used for model development, while data from 46 (22.8%) patients from a multinational institution were used for external validation. A total of 21,074 kidney and stone contour-annotated CT images were trained with the ResNet-18 Mask R-convolutional neural network algorithm. Finally, this model was concatenated with demographic and clinical data as a fully connected layer for stone classification. Our model was 100% sensitive in detecting kidney stones in each patient, and the delineation of kidney and stone contours was precise within clinically acceptable ranges. The development model provided an accuracy of 99.9%, with 100.0% sensitivity and 98.9% specificity, in distinguishing UA from non-UA stones. On external validation, the model performed with an accuracy of 97.1%, with 89.4% sensitivity and 98.6% specificity. SHAP plots revealed stone density, diabetes mellitus, and urinary pH as the most important features for classification. Our ML-based model accurately identified and delineated kidney stones and classified UA stones from non-UA stones with the highest predictive accuracy reported to date. Our model can be reliably used to select candidates for an earlier-directed alkalization therapy.

Toward Precision Medicine: Development and Validation of A Machine Learning Based Decision Support System for Optimal Sequencing in Castration-Resistant Prostate Cancer.

INTRODUCTION: Selecting a patient-specific sequencing strategy to maximize survival outcomes is a clinically unmet need for patients with castration-resistant prostate cancer (CRPC). We developed and validated an artificial intelligence-based decision support system (DSS) to guide optimal sequencing strategy selection. PATIENTS AND METHODS: Clinicopathological data of 46 covariates were retrospectively collected from 801 patients diagnosed with CRPC at 2 high-volume institutions between February 2004 and March 2021. Cox-proportional hazards regression survival (Cox) modeling in extreme gradient boosting (XGB) was used to perform survival analysis for cancer-specific mortality (CSM) and overall mortality (OM) according to the use of abiraterone acetate, cabazitaxel, docetaxel, and enzalutamide. The models were further stratified into first-, second-, and third-line models that each provided CSM and OM estimates for each line of treatment. The performances of the XGB models were compared with those of the Cox models and random survival forest (RSF) models in terms of Harrell's C-index. RESULTS: The XGB models showed greater predictive performance for CSM and OM compared to the RSF and Cox models. C-indices of 0.827, 0.807, and 0.748 were achieved for CSM in the first-, second-, and third-lines of treatment, respectively, while C-indices of 0.822, 0.813, and 0.729 were achieved for OM regarding each line of treatment, respectively. An online DSS was developed to provide visualization of individualized survival outcomes according to each line of sequencing strategy. CONCLUSION: Our DSS can be used in clinical practice by physicians and patients as a visualized tool to guide the sequencing strategy of CRPC agents.