Confounding factors in image analysis

This chapter explores confounding factors in the MRI pipeline from patient presentation to diagnosis for their influences on quantitative parameters. It categorizes confounders into patient-related, acquisition-related, and reader-dependent factors and systematically investigates their impact on annotations, clinical parameters (CPs), and outcomes. Methods for identifying, tracing, and mitigating confounders are discussed, with a focus on comparison strategies for reader-, sequence-, and acquisition- influences, as well as statistical process control. Practical examples illustrate the challenges of confounder management, including the effects of artifacts, the relevance of reader competency in parametric mapping, and the effects of mathematical models of annotations on CPs. Recommendations are provided for improving robustness in clinical practice, while additionally providing an outlook into novel methods for controlling and standardizing quantitative parameter differences across confounders. This chapter highlights the critical importance of addressing confounding factors to ensure reliable, reproducible, and generalizable imaging results.