Prediction of 5-year survival in soft tissue leiomyosarcoma using a machine learning model algorithm.

Area of Science:

• Oncology
• Machine Learning in Medicine
• Cancer Prognostics

Background:

• Leiomyosarcoma (LMS) has a poor prognosis among soft tissue sarcomas.
• Accurate survival prediction is crucial for patient management.

Purpose of the Study:

• To develop and validate machine learning models for 5-year survival prediction in appendicular or truncal LMS.
• To identify the best-performing model for clinical utility.

Main Methods:

• Utilized the SEER database for model development and internal validation.
• Externally validated five machine learning algorithms using an institutional database.
• Assessed model performance using Area Under the Receiver Operating Characteristic Curve (AUC) and Brier score.

Main Results:

• Developed five machine learning algorithms for LMS survival prediction.
• Models demonstrated excellent calibration (AUC 0.84-0.85, Brier score 0.15-0.16).
• Elastic-Net Penalized Logistic Regression showed superior performance, with AUC 0.85 and Brier score 0.15 on external validation.

Conclusions:

• Successfully developed and validated machine learning models for 5-year LMS survival prediction.
• Elastic-Net Penalized Logistic Regression is a robust model for predicting survival in LMS patients.
• This model can aid in clinical decision-making and patient counseling.