Statistically Guided Improvement of Medical Imaging Analysis

Abstract

Quantitative biomedical imaging is a rapidly evolving field, integrating novel biotechnologies with advanced quantitative methods to unlock new scientific insights. Alongside these developments come challenges, particularly with emerging imaging modalities. My dissertation develops statistical methods to address cutting-edge problems in biomedical imaging through two projects. The first project introduces GammaGateR, an R package implementing a statistical framework tailored for semi-automated marker gating in high-dimensional images. The second project investigates statistical inference for AI-based image synthesis, addressing the challenges posed by missing uncertainty and clinical information, and proposing practical ad hoc solutions. Together, these contributions enhance quantitative image analysis, providing tools and insights to improve both methodological rigor and practical applications in biomedical research.