Ultra-high dimensional confounder selection algorithms comparison with application to radiomics data

TL;DR

This paper compares machine learning algorithms for causal inference in ultra-high dimensional radiomics data, demonstrating that SIS + GOAL outperforms other methods in cancer imaging studies.

Contribution

It introduces an extension of GOAL and OAL algorithms for ultra-high dimensional data and evaluates their performance in radiomics applications.

Findings

SIS + GOAL was identified as the best variable selection method.

The study demonstrates the effectiveness of causal inference algorithms in radiomics.

SIS + OAL and CBS performed less optimally in the tested datasets.

Abstract

Radiomics is an emerging area of medical imaging data analysis particularly for cancer. It involves the conversion of digital medical images into mineable ultra-high dimensional data. Machine learning algorithms are widely used in radiomics data analysis to develop powerful decision support model to improve precision in diagnosis, assessment of prognosis and prediction of therapy response. However, machine learning algorithms for causal inference have not been previously employed in radiomics analysis. In this paper, we evaluate the value of machine learning algorithms for causal inference in radiomics. We select three recent competitive variable selection algorithms for causal inference: outcome-adaptive lasso (OAL), generalized outcome-adaptive lasso (GOAL) and causal ball screening (CBS). We used a sure independence screening procedure to propose an extension of GOAL and OAL for ultra-high dimensional data, SIS + GOAL and SIS + OAL. We compared SIS + GOAL, SIS + OAL and CBS using simulation study and two radiomics datasets in cancer, osteosarcoma and gliosarcoma. The two radiomics studies and the simulation study identified SIS + GOAL as the optimal variable selection algorithm.