Selected work

Representative Papers

Expert Consensus on Criteria for the Automated Assessment of Laparoscopic Camera Navigation

Jun 22, 2026

This study addresses the longstanding reliance on subjective and inefficient manual scoring in assessing laparoscopic camera navigation skills, which lacks standardized and scalable objective metrics. The authors propose a novel evaluation taxonomy comprising 14 key elements, aligning clinical importance—established through expert consensus—with technical readiness of computer vision methods via a “clinical importance–technical readiness” matrix to prioritize automation targets. Through Likert-scale surveys, expert-based skill rankings, and computer vision–derived automated measurements, validated across 23 surgeons, the study identifies high-priority metrics such as field-of-view coverage, focus quality, and instrument centering. These metrics jointly satisfy clinical relevance and technical feasibility, establishing a practical framework for AI-driven surgical training assessment.

0 citationsRead paper

Empirical prior distributions for treatment-by-subgroup interaction heterogeneity in random-effects meta-analysis

Jun 22, 2026

This study addresses the imprecise inference in subgroup interaction meta-analyses under sparse data, which stems from the absence of empirical prior distributions tailored to interaction heterogeneity. Leveraging over 3,000 interaction meta-analyses from the Cochrane Database of Systematic Reviews, we construct the first treatment-by-subgroup interaction–specific empirical prior distribution, revealing that such interaction heterogeneity is typically substantially smaller than that of overall treatment effects. By integrating a Bayesian random-effects model with large-scale data mining and predictive prior derivation, the proposed prior markedly improves estimation accuracy in sparse-data settings compared to conventional heterogeneity priors, thereby offering a more reliable evidentiary foundation for subgroup analyses.

0 citationsRead paper

Mitigating Measurement-Induced Training Instability in Hybrid Quantum Neural Networks for Protein Classification

Jun 21, 2026

This work identifies and formally names a previously unrecognized issue in hybrid quantum neural networks—“measurement-induced logit contraction”—where quantum measurement outputs, constrained to the interval [−1, 1], diminish the sensitivity of cross-entropy loss to logit differences, leading to vanishing gradients and unstable training. To address this, the authors propose a circuit-agnostic, learnable Quantum Measurement Temperature (QMT) mechanism that adaptively scales measurement outputs to enhance loss sensitivity without altering the underlying quantum circuit architecture. Experimental results demonstrate that QMT substantially improves logit separation, gradient magnitude, and training stability, yielding higher classification accuracy on both fluorescence microscopy images and a six-class Fashion-MNIST benchmark.

0 citationsRead paper

FederatedRSF : Federated Random Survival Forests for Partially Overlapping Medical Data

May 21, 2026

This study addresses the challenge of multicenter survival prediction, where patient-level clinical and genomic data cannot be shared due to privacy regulations and feature spaces across centers only partially overlap. The work proposes a federated random survival forest method that accommodates heterogeneous features by training survival trees locally and aggregating only those trees whose features are compatible across sites, thereby enabling collaborative modeling without exchanging raw data. Evaluated on simulated multicenter settings using the GBSG2 breast cancer dataset through repeated cross-validation, the proposed approach achieves performance comparable to centralized training as measured by Harrell’s C-index, effectively balancing privacy preservation with predictive efficacy.

0 citationsRead paper

Identifying the potential of sample overlap in evidence synthesis of observational studies

Feb 24, 2026

This study addresses the challenge of sample overlap in evidence synthesis from observational studies, which can introduce substantial bias—particularly when individual-level identifiers are unavailable to detect or correct such overlap. To overcome this limitation, the authors propose a novel method grounded in set theory that requires no individual participant data. By encoding the ranges of multiple carefully selected sample characteristics, the approach constructs an index quantifying the degree of sample overlap, enabling inference of overlapping samples and identification of the largest non-overlapping subset. This method fills a critical gap in the secondary use of real-world evidence, where handling sample overlap has been underexplored. Its validity and flexibility are demonstrated across several empirical case studies, significantly enhancing the credibility of synthesized evidence.

0 citationsRead paper

Recent publications

Latest Papers

Expert Consensus on Criteria for the Automated Assessment of Laparoscopic Camera Navigation

Jun 22, 2026

This study addresses the longstanding reliance on subjective and inefficient manual scoring in assessing laparoscopic camera navigation skills, which lacks standardized and scalable objective metrics. The authors propose a novel evaluation taxonomy comprising 14 key elements, aligning clinical importance—established through expert consensus—with technical readiness of computer vision methods via a “clinical importance–technical readiness” matrix to prioritize automation targets. Through Likert-scale surveys, expert-based skill rankings, and computer vision–derived automated measurements, validated across 23 surgeons, the study identifies high-priority metrics such as field-of-view coverage, focus quality, and instrument centering. These metrics jointly satisfy clinical relevance and technical feasibility, establishing a practical framework for AI-driven surgical training assessment.

0 citationsRead paper

Empirical prior distributions for treatment-by-subgroup interaction heterogeneity in random-effects meta-analysis

Jun 22, 2026

This study addresses the imprecise inference in subgroup interaction meta-analyses under sparse data, which stems from the absence of empirical prior distributions tailored to interaction heterogeneity. Leveraging over 3,000 interaction meta-analyses from the Cochrane Database of Systematic Reviews, we construct the first treatment-by-subgroup interaction–specific empirical prior distribution, revealing that such interaction heterogeneity is typically substantially smaller than that of overall treatment effects. By integrating a Bayesian random-effects model with large-scale data mining and predictive prior derivation, the proposed prior markedly improves estimation accuracy in sparse-data settings compared to conventional heterogeneity priors, thereby offering a more reliable evidentiary foundation for subgroup analyses.

0 citationsRead paper

Mitigating Measurement-Induced Training Instability in Hybrid Quantum Neural Networks for Protein Classification

Jun 21, 2026

This work identifies and formally names a previously unrecognized issue in hybrid quantum neural networks—“measurement-induced logit contraction”—where quantum measurement outputs, constrained to the interval [−1, 1], diminish the sensitivity of cross-entropy loss to logit differences, leading to vanishing gradients and unstable training. To address this, the authors propose a circuit-agnostic, learnable Quantum Measurement Temperature (QMT) mechanism that adaptively scales measurement outputs to enhance loss sensitivity without altering the underlying quantum circuit architecture. Experimental results demonstrate that QMT substantially improves logit separation, gradient magnitude, and training stability, yielding higher classification accuracy on both fluorescence microscopy images and a six-class Fashion-MNIST benchmark.

0 citationsRead paper

FederatedRSF : Federated Random Survival Forests for Partially Overlapping Medical Data

May 21, 2026

This study addresses the challenge of multicenter survival prediction, where patient-level clinical and genomic data cannot be shared due to privacy regulations and feature spaces across centers only partially overlap. The work proposes a federated random survival forest method that accommodates heterogeneous features by training survival trees locally and aggregating only those trees whose features are compatible across sites, thereby enabling collaborative modeling without exchanging raw data. Evaluated on simulated multicenter settings using the GBSG2 breast cancer dataset through repeated cross-validation, the proposed approach achieves performance comparable to centralized training as measured by Harrell’s C-index, effectively balancing privacy preservation with predictive efficacy.

0 citationsRead paper

Identifying the potential of sample overlap in evidence synthesis of observational studies

Feb 24, 2026

This study addresses the challenge of sample overlap in evidence synthesis from observational studies, which can introduce substantial bias—particularly when individual-level identifiers are unavailable to detect or correct such overlap. To overcome this limitation, the authors propose a novel method grounded in set theory that requires no individual participant data. By encoding the ranges of multiple carefully selected sample characteristics, the approach constructs an index quantifying the degree of sample overlap, enabling inference of overlapping samples and identification of the largest non-overlapping subset. This method fills a critical gap in the secondary use of real-world evidence, where handling sample overlap has been underexplored. Its validity and flexibility are demonstrated across several empirical case studies, significantly enhancing the credibility of synthesized evidence.

0 citationsRead paper