Development of a machine learning-based model for prognostic prediction in melanoma

Area of Science:

• Oncology
• Bioinformatics
• Machine Learning

Background:

• Melanoma is an aggressive skin cancer with a poor prognosis, making survival prediction critical.
• Accurate prognostic tools are needed to support clinical decision-making and improve patient outcomes.

Purpose of the Study:

• To apply and evaluate five machine learning models for predicting melanoma patient survival rates.
• To identify the most accurate model for improving prognostic assessment and clinical utility.

Main Methods:

• Utilized data from 4,875 cutaneous melanoma patients from the SEER database.
• Applied and compared Random Forest, Decision Tree, XGBoost, CatBoost, and LightGBM models.
• Evaluated model performance using Area Under the ROC Curve (AUC), confusion matrix, calibration curves, and Decision Curve Analysis (DCA).

Main Results:

• The CatBoost model demonstrated the best performance and stability across 1-, 3-, and 5-year survival predictions.
• Achieved AUC values of 0.7577, 0.7595, and 0.7557 for 1-, 3-, and 5-year survival, respectively.
• Decision Curve Analysis confirmed the clinical utility of the CatBoost model.

Conclusions:

• The CatBoost model is a practical and accurate tool for melanoma prognosis assessment.
• This machine learning approach supports individualized clinical decision-making and may enhance early intervention strategies.
• The model's robust generalization capabilities offer reliable predictive insights for melanoma patient survival.