Research

My work develops statistical inference methods and software that can be used for scientific insight in medical imaging, primarily motivated by magnetic resonance imaging studies in humans. In collaborative research, I utilize statistical tools such as generalized additive models, random effects models, machine learning, semiparametric methods, and estimating equations. My statistical research focuses on robust inference procedures for high-dimensional imaging data using semiparametric methods. Lately, I am particularly interested in the use of robust effect sizes, semiparametric theory, and replicability in medical imaging.

The research content below might not be up-to-date. For my latest interests, please see my Google Scholar page.

Effect sizes in neuroimaging

• I am involved in the development and application of the robust effect size index (RESI).
• Kaidi Kang’s paper defined a cross sectional RESI (CS-RESI) for longitudinal models that makes effect sizes comparable across cross-sectional and longitudinal studies. His work investigated ways to increase effect sizes in BWAS studies. The preprint is on Biorxiv.
• The interactive version of a figure from our 2020 OHBM paper comparing effect size thresholding to \(p\)-value thresholding for neuroimaging inference.
• Megan Jones wrote the sofware and paper for the RESI R package available on CRAN and in press with Journal of Statistical Software. The preprint is available on ArXiv.

Multiplexed imaging

• Ruby Xiong developed a semi-automated approach for single-cell marker gating/phenotyping for multiplexed imaging data. The preprint is on Biorxiv.
• Harsimran Kaur developed a tool to define consensus tissue domains across a large number of slides. The preprint is on Biorxiv.