Selected work

Representative Papers

Foundation Model-Enabled Efficient Data Sampling (FEEDS): A label-efficient training strategy for pan-cancer, multi-tracer PET/CT datasets

Aug 11, 2026

This work addresses the performance and generalization bottlenecks in whole-body lesion segmentation caused by scarce annotations, proposing FEEDS—a single-step, efficient training paradigm grounded in vision foundation model embeddings. FEEDS selects informative and diverse samples from large-scale unlabeled PET/CT data in a single pass for expert annotation, eliminating the need for iterative pseudo-labeling or active learning loops, thereby substantially reducing both annotation and computational costs. Using only 30% of the annotated data, the method achieves segmentation performance on par with fully supervised models across multiple datasets—including AutoPET-III, DeepPSMA, and an internal multicenter, multi-tracer cohort—demonstrating significantly superior generalization compared to random sampling and semi-supervised baselines.

0 citationsRead paper

Constructing external comparator groups via transportability in mean or in effect measure

Apr 21, 2026

This study addresses the challenge of constructing high-quality external control arms for target populations using multi-source data to reliably estimate causal effects of alternative treatments. To this end, the authors propose two identification strategies—one based on the transportability of potential outcome means and the other on the transportability of effect measures—and develop a semiparametric, doubly robust augmented weighting estimator. This estimator integrates models for trial participation probability, treatment assignment probability, and conditional outcome mean, maintaining robustness under partial model misspecification. Theoretical analysis establishes its asymptotic efficiency, while simulation studies demonstrate superior finite-sample performance compared to methods relying solely on modeling or weighting. The approach is successfully applied to the ACCEPT and PHOENIX 1 trials, yielding reliable assessments of efficacy differences among biologic therapies for psoriasis.

0 citationsRead paper

Using functional information for binary classifications

Dec 03, 2025

This paper addresses binary classification for continuous-time functional data. We propose a probabilistic binary classification (PBC) criterion grounded in inter-functional distances, along with a fully nonparametric estimation procedure. Unlike conventional paradigms relying on ordered scalar labels, PBC directly quantifies similarity between an individual’s trajectory and the functional mean of the positive group—bypassing structural assumptions or dimensionality reduction on functional biomarkers. The method integrates functional data analysis, nonparametric kernel estimation, and Monte Carlo simulation, implemented in R. Simulation studies and real-data analyses demonstrate that PBC exhibits strong robustness under moderate sample sizes and achieves significantly higher classification accuracy than state-of-the-art competitors—particularly in challenging settings involving high-dimensional, nonstationary, and small-sample functional biomarkers.

0 citationsRead paper

Recent publications

Latest Papers

Foundation Model-Enabled Efficient Data Sampling (FEEDS): A label-efficient training strategy for pan-cancer, multi-tracer PET/CT datasets

Aug 11, 2026

This work addresses the performance and generalization bottlenecks in whole-body lesion segmentation caused by scarce annotations, proposing FEEDS—a single-step, efficient training paradigm grounded in vision foundation model embeddings. FEEDS selects informative and diverse samples from large-scale unlabeled PET/CT data in a single pass for expert annotation, eliminating the need for iterative pseudo-labeling or active learning loops, thereby substantially reducing both annotation and computational costs. Using only 30% of the annotated data, the method achieves segmentation performance on par with fully supervised models across multiple datasets—including AutoPET-III, DeepPSMA, and an internal multicenter, multi-tracer cohort—demonstrating significantly superior generalization compared to random sampling and semi-supervised baselines.

0 citationsRead paper

Constructing external comparator groups via transportability in mean or in effect measure

Apr 21, 2026

This study addresses the challenge of constructing high-quality external control arms for target populations using multi-source data to reliably estimate causal effects of alternative treatments. To this end, the authors propose two identification strategies—one based on the transportability of potential outcome means and the other on the transportability of effect measures—and develop a semiparametric, doubly robust augmented weighting estimator. This estimator integrates models for trial participation probability, treatment assignment probability, and conditional outcome mean, maintaining robustness under partial model misspecification. Theoretical analysis establishes its asymptotic efficiency, while simulation studies demonstrate superior finite-sample performance compared to methods relying solely on modeling or weighting. The approach is successfully applied to the ACCEPT and PHOENIX 1 trials, yielding reliable assessments of efficacy differences among biologic therapies for psoriasis.

0 citationsRead paper

Using functional information for binary classifications

Dec 03, 2025

This paper addresses binary classification for continuous-time functional data. We propose a probabilistic binary classification (PBC) criterion grounded in inter-functional distances, along with a fully nonparametric estimation procedure. Unlike conventional paradigms relying on ordered scalar labels, PBC directly quantifies similarity between an individual’s trajectory and the functional mean of the positive group—bypassing structural assumptions or dimensionality reduction on functional biomarkers. The method integrates functional data analysis, nonparametric kernel estimation, and Monte Carlo simulation, implemented in R. Simulation studies and real-data analyses demonstrate that PBC exhibits strong robustness under moderate sample sizes and achieves significantly higher classification accuracy than state-of-the-art competitors—particularly in challenging settings involving high-dimensional, nonstationary, and small-sample functional biomarkers.

0 citationsRead paper