RCC-Supporter: supporting renal cell carcinoma treatment decision-making using machine learning.

Area of Science:

• Oncology
• Medical Informatics
• Artificial Intelligence

Background:

• Renal cell carcinoma (RCC) incidence is rising globally, particularly in Asia.
• Diverse treatment options for RCC depend on clinical stage and histology.
• Personalized treatment strategies are crucial for effective RCC management.

Purpose of the Study:

• To develop a machine learning (ML)-based clinical decision-support system (CDSS) for personalized RCC treatment recommendations.
• To tailor treatment decisions to individual patient health conditions and clinical situations.

Main Methods:

• Utilized real-world medical data from 1,867 RCC patients treated between 2008 and 2021.
• Categorized patients into Surveillance, Surgery, and Chemotherapy groups.
• Applied feature selection to identify significant clinical factors from 2,058 features.
• Implemented and evaluated Decision Tree, Random Forest, and Gradient Boosting Machine (GBM) algorithms.

Main Results:

• The Gradient Boosting Machine (GBM) algorithm achieved 95% accuracy (95% CI, 92-98%) using 100 and 150 features.
• GBM with 100/150 features outperformed expert-selected features (93% accuracy).
• The developed system, 'RCC-Supporter', shows high performance in predicting personalized treatment.

Conclusions:

• A preliminary personalized treatment decision-support system (TDSS) named 'RCC-Supporter' was developed using ML algorithms.
• The study demonstrates the feasibility of ML-based CDSS for RCC treatment decisions in clinical practice.
• This approach offers a promising tool for optimizing patient care and outcomes in renal cell carcinoma.