Using machine learning to predict lymph node metastasis in patients with renal cell carcinoma: A population-based...

Area of Science:

• Oncology
• Medical Informatics
• Machine Learning

Background:

• Lymph node (LN) metastasis in renal cell carcinoma (RCC) is a key indicator of poor prognosis and distant metastasis.
• Accurate prediction of LN metastasis is vital for personalized treatment strategies and surgical decisions in RCC patients.
• There is a need for a robust prediction model to assess the risk of LN metastasis in RCC.

Purpose of the Study:

• To develop and validate a machine learning model for predicting lymph node metastasis in renal cell carcinoma (RCC).
• To create an accessible online risk calculator for assessing LN metastasis risk in RCC patients.
• To evaluate the impact of predicted LN metastasis on overall survival in RCC.

Main Methods:

• Utilized partial data from the SEER database and external validation data from 492 RCC patients.
• Screened eight indicators for LN metastasis risk and established six machine learning (ML) classifiers.
• Developed an online risk calculator based on the optimal Extreme Gradient-Boosting (XGB) model and performed survival analysis.

Main Results:

• The XGB model demonstrated superior performance in both internal and external validation, with high accuracy, sensitivity, and specificity.
• Achieved an area under the curve (AUC) of 0.930 internally and 0.958 externally, indicating excellent predictive ability.
• Survival analysis revealed significantly shorter survival for patients with predicted N1 tumors (p < 0.0001).

Conclusions:

• Integrating ML algorithms with clinical data effectively predicts LN metastasis in RCC patients.
• A freely available online calculator based on the XGB model has been developed for clinical use.
• The study highlights the clinical utility of ML in improving the management of RCC patients.