Survival analysis with semi-supervised predictive clustering trees.

Area of Science:

• Biostatistics
• Machine Learning
• Clinical Informatics

Background:

• Clinical studies often involve time-to-event data with censored observations.
• Accurate survival prediction is crucial for personalized medicine.

Purpose of the Study:

• To develop a novel approach for time-to-event prediction using semi-supervised learning.
• To model censored observations as partially labeled examples for improved risk prediction.

Main Methods:

• Framing time-to-event prediction as a multi-target regression task.
• Applying semi-supervised learning, specifically semi-supervised predictive clustering trees and ensembles.
• Evaluating performance on eleven real-life datasets.

Main Results:

• The proposed method achieved superior or equivalent predictive performance compared to three competitor methods.
• Generated smaller models than random survival forests.
• Demonstrated informative feature selection for survival prediction in amyotrophic lateral sclerosis patients.

Conclusions:

• Semi-supervised learning offers a powerful framework for time-to-event prediction with censored data.
• The proposed method provides an effective and efficient alternative to existing survival analysis techniques.
• The approach aids in understanding key predictors for patient survival.