Using Real-World Data for Machine-Learning Algorithms to Predict the Treatment Response in Advanced Melanoma: A Pilot...

Area of Science:

• Oncology
• Medical Informatics
• Artificial Intelligence

Background:

• Real-world data (RWD) is crucial for oncology clinical decision-making and personalized treatment.
• Advanced melanoma patients show variable responses to targeted therapy and immunotherapy, necessitating personalized approaches.

Purpose of the Study:

• To predict clinical outcomes in advanced melanoma using machine learning (ML) and RWD.
• To apply explainable artificial intelligence (XAI) for understanding individual treatment predictions.

Main Methods:

• Developed and validated four ML models using RWD from 239 melanoma patients.
• Incorporated ML for predictive modeling and XAI for model interpretability.
• Evaluated model performance in predicting 2-year survival.

Main Results:

• ML models achieved an AUC >80% and accuracy >74% in predicting 2-year survival.
• The random forest model demonstrated the highest performance with an AUC of 0.85.
• XAI provided insights into individual predictions, enhancing model trust and clinical relevance.

Conclusions:

• Integrated RWD and ML for predicting melanoma patient outcomes, demonstrating proof-of-concept.
• XAI enhances the usability and trustworthiness of ML models in clinical settings.
• Future research with advanced AI can further improve prognostic and predictive models for melanoma.