A Support vector machine-based mixture cure model for mixed case interval censored data

Area of Science:

• Biostatistics
• Machine Learning
• Survival Analysis

Background:

• Mixed case interval censored (MCIC) data presents unique challenges in statistical analysis.
• Identifying a 'cured' subgroup, where individuals are never susceptible to the event, is crucial for accurate modeling.

Purpose of the Study:

• To develop a novel semi-parametric two-component model for analyzing MCIC data with a cured subgroup.
• To integrate a support vector machine (SVM) for enhanced cure probability modeling and a Cox proportional hazards structure for survival distribution analysis.
• To address the limitations of traditional generalized linear models in capturing complex covariate effects within MCIC data.

Main Methods:

• A semi-parametric two-component model combining SVM for cure probability and Cox proportional hazards for uncured survival.
• Development of an expectation maximization algorithm for parameter estimation.
• Simulation studies to evaluate model performance and superiority.

Main Results:

• The proposed SVM-based model demonstrates superior performance compared to traditional methods in simulation studies.
• Successful application of the model to NASA's Hypobaric Decompression Sickness Data.
• The model effectively captures complex covariate effects and handles interval censored data with a cured subgroup.

Conclusions:

• The novel SVM-based semi-parametric model offers a powerful and flexible approach for analyzing MCIC data with cured subgroups.
• This work represents the first application of machine learning algorithms to MCIC data analysis in the presence of a cured population.
• The model provides improved accuracy and interpretability for survival data analysis in various scientific fields.