Motivation: Measurement error in quantitative imaging markers (QIMs) biases parameter estimation and reduces statistical power, limiting the reliability of clinical imaging studies. Goal(s): Demonstrate the importance of modeling QIM measurement error using regression calibration to improve estimation of associations between QIMs and clinical outcomes. Approach: We provide a framework for measurement error calibration in medical imaging studies, which we demonstrate in a clinical study for epilepsy seizure onset zone detection. Results: Correcting for measurement error significantly improved the estimation of associations, enhanced statistical power, corrected bias in regression analyses, and informed appropriate sample size estimation for clinical imaging studies. Impact: We demonstrate how measurement error modeling significantly enhances clinical imaging study design and analysis by improving parameter estimation, statistical power, and study reliability. Our methodology is particularly valuable for multicenter and longitudinal studies where measurement variability is a key concern.
Wang et al. (Tue,) studied this question.