Renal Cell Carcinoma: Promise and Pitfalls of AI in Research.

What is it about?

The authors thank Khene et al. for their comments and clarify methodological points in their study. They followed the STREAM-URO guidelines and used an external validation dataset. The missing data in the R.E.N.A.L. nephrometry score predictor was acceptable. The decision curve showed higher net reduction in interventions per 100 patients. The ML model was trained on the training dataset and hyperparameters were determined using a Bayesian optimization approach. Khene et al. suggested using deep neural networks and Shapley Additive exPlanations, but these were inappropriate for this study.

Featured Image

Why is it important?

The research is important as it presents a novel predictive signature for pT3a upstaging in patients with renal cell carcinoma. This has clinical implications as it allows for better patient management and treatment decisions. Key Takeaways: 1. The study follows the STREAM-URO guidelines and uses an external validation dataset to ensure the robustness of the predictive signature. 2. The model used in the study is trained exclusively on the training dataset and then locked and evaluated on the testing dataset, ensuring an unbiased estimate of the predictive performance. 3. The authors address comments raised by Khene et al. and provide a decision curve showing higher net reduction in interventions per 100 patients when using the model for decision-making across relevant threshold probabilities.