Regularized Buckley-James method for right-censored outcomes with block-missing multimodal covariates

Area of Science:

• Medical research
• Biostatistics
• Neuroimaging analysis

Background:

• High-dimensional data with censored outcomes are common in medical research.
• The regularized Buckley-James estimator is useful for predictive modeling and variable selection in such data.
• Analyzing multimodal neuroimaging data with missing covariates and censored outcomes presents unique challenges.

Purpose of the Study:

• To develop a statistical method for parameter estimation and variable selection in semiparametric accelerated failure time models.
• To address high-dimensional block-missing multimodal neuroimaging data with censored outcomes.
• To propose a penalized Buckley-James method capable of handling both missing covariates and censored data.

Main Methods:

• A penalized Buckley-James method is proposed to simultaneously manage block-wise missing covariates and censored outcomes.
• The method incorporates variable selection capabilities within the analysis framework.
• Statistical simulations were conducted to evaluate the performance of the proposed method.

Main Results:

• The proposed penalized Buckley-James method demonstrated effectiveness in simulations.
• The method was successfully applied to a multimodal neuroimaging dataset.
• Meaningful results were obtained from the analysis of the neuroimaging data.

Conclusions:

• The penalized Buckley-James method provides a robust approach for analyzing high-dimensional neuroimaging data with missingness and censoring.
• This method facilitates accurate parameter estimation and effective variable selection in complex medical datasets.
• The findings suggest the utility of the proposed method in advancing medical research and neuroimaging analysis.