Selected work

Representative Papers

Algorithmic statistics of retinal images

Aug 06, 2026

This study addresses the issue of physiologically irrelevant systematic distortions introduced by existing non-metric methods when analyzing three-dimensional retinal OCT images, which can compromise the assessment of disease progression. The authors propose a metric learning framework that, for the first time, integrates Normalized Compression Distance (NCD) with anisotropic structure-enhancement filtering to construct interpretable and metrically consistent Normalized Compression Vector (NCV) representations. This approach effectively reveals category-dependent statistical distortions potentially induced by non-metric embeddings. Experimental results demonstrate that NCV predicts visual field functional changes with an error of approximately 0.5 dB, outperforming current non-metric deep learning methods. Furthermore, the framework’s capacity to quantify and visualize structural differences is validated in both human glaucoma patients and non-human primate models.

0 citationsRead paper

An Actuarial Cost and Revenue Model for Helicopter Emergency Medical Services: Estimating Population-Based Coverage and Sustainability Thresholds

Jun 11, 2026

This study addresses the financial sustainability challenges of Helicopter Emergency Medical Services (HEMS) by developing a transparent, replicable two-stage evaluation framework that integrates actuarial methodologies with real-world insurance reimbursement data. The framework comprises a cost-accounting module and a revenue-prediction model, enhanced through Monte Carlo simulation (10,000 iterations) and multi-scenario sensitivity analyses. Under a baseline scenario assuming 50% commercial insurance reimbursement of billed charges and 24/7 staffing, the model identifies a breakeven threshold of 184 annual transports. However, this requirement surges to over 1,000 missions if reimbursement is limited to Medicare rates or if personnel costs double. For the first time, the model quantifies the minimum population coverage threshold necessary for HEMS viability, offering policymakers and healthcare planners an evidence-based tool for strategic decision-making and resource allocation.

0 citationsRead paper

Reinforcement Learning Improves LLM Accuracy and Reasoning in Disease Classification from Radiology Reports

Apr 21, 2026

This work addresses the trade-off between classification accuracy and reasoning capability in radiology report disease classification, where supervised fine-tuning often improves accuracy at the expense of model interpretability. To reconcile this, the authors propose a two-stage approach: first, a lightweight large language model is fine-tuned with disease labels under supervision; subsequently, Group Relative Policy Optimization (GRPO) is applied to refine model outputs without requiring explicit reasoning annotations. This study presents the first application of GRPO to radiology text classification. Evaluated on three datasets annotated by radiologists, the method not only significantly outperforms baseline models in classification performance but also concurrently enhances reasoning recall and content comprehensiveness, achieving a synergistic improvement in both accuracy and reasoning quality.

0 citationsRead paper

Generative Large Language Models Trained for Detecting Errors in Radiology Reports

Apr 06, 2025

This study addresses four clinically critical semantic errors in radiology reports—negation, laterality, temporal progression, and transcription—by constructing the first dual-source annotated dataset integrating GPT-4–synthesized erroneous samples with real MIMIC-CXR reports. We propose an LLM-based automated error detection framework and demonstrate, for the first time, that Llama-3-70B-Instruct achieves high zero-shot performance on clinical semantic error identification. Supervised fine-tuning further improves accuracy, yielding an overall F1-score of 0.780. A double-blind evaluation by radiologists on 200 flagged instances confirmed 163 as genuine errors, achieving an 81.5% clinical acceptance rate. This work establishes a novel LLM-driven paradigm for radiology report quality control and provides a reproducible, verifiable methodological foundation for automated clinical text quality assurance.

0 citationsRead paper

Recent publications

Latest Papers

Algorithmic statistics of retinal images

Aug 06, 2026

This study addresses the issue of physiologically irrelevant systematic distortions introduced by existing non-metric methods when analyzing three-dimensional retinal OCT images, which can compromise the assessment of disease progression. The authors propose a metric learning framework that, for the first time, integrates Normalized Compression Distance (NCD) with anisotropic structure-enhancement filtering to construct interpretable and metrically consistent Normalized Compression Vector (NCV) representations. This approach effectively reveals category-dependent statistical distortions potentially induced by non-metric embeddings. Experimental results demonstrate that NCV predicts visual field functional changes with an error of approximately 0.5 dB, outperforming current non-metric deep learning methods. Furthermore, the framework’s capacity to quantify and visualize structural differences is validated in both human glaucoma patients and non-human primate models.

0 citationsRead paper

An Actuarial Cost and Revenue Model for Helicopter Emergency Medical Services: Estimating Population-Based Coverage and Sustainability Thresholds

Jun 11, 2026

This study addresses the financial sustainability challenges of Helicopter Emergency Medical Services (HEMS) by developing a transparent, replicable two-stage evaluation framework that integrates actuarial methodologies with real-world insurance reimbursement data. The framework comprises a cost-accounting module and a revenue-prediction model, enhanced through Monte Carlo simulation (10,000 iterations) and multi-scenario sensitivity analyses. Under a baseline scenario assuming 50% commercial insurance reimbursement of billed charges and 24/7 staffing, the model identifies a breakeven threshold of 184 annual transports. However, this requirement surges to over 1,000 missions if reimbursement is limited to Medicare rates or if personnel costs double. For the first time, the model quantifies the minimum population coverage threshold necessary for HEMS viability, offering policymakers and healthcare planners an evidence-based tool for strategic decision-making and resource allocation.

0 citationsRead paper

Reinforcement Learning Improves LLM Accuracy and Reasoning in Disease Classification from Radiology Reports

Apr 21, 2026

This work addresses the trade-off between classification accuracy and reasoning capability in radiology report disease classification, where supervised fine-tuning often improves accuracy at the expense of model interpretability. To reconcile this, the authors propose a two-stage approach: first, a lightweight large language model is fine-tuned with disease labels under supervision; subsequently, Group Relative Policy Optimization (GRPO) is applied to refine model outputs without requiring explicit reasoning annotations. This study presents the first application of GRPO to radiology text classification. Evaluated on three datasets annotated by radiologists, the method not only significantly outperforms baseline models in classification performance but also concurrently enhances reasoning recall and content comprehensiveness, achieving a synergistic improvement in both accuracy and reasoning quality.

0 citationsRead paper

Generative Large Language Models Trained for Detecting Errors in Radiology Reports

Apr 06, 2025

This study addresses four clinically critical semantic errors in radiology reports—negation, laterality, temporal progression, and transcription—by constructing the first dual-source annotated dataset integrating GPT-4–synthesized erroneous samples with real MIMIC-CXR reports. We propose an LLM-based automated error detection framework and demonstrate, for the first time, that Llama-3-70B-Instruct achieves high zero-shot performance on clinical semantic error identification. Supervised fine-tuning further improves accuracy, yielding an overall F1-score of 0.780. A double-blind evaluation by radiologists on 200 flagged instances confirmed 163 as genuine errors, achieving an 81.5% clinical acceptance rate. This work establishes a novel LLM-driven paradigm for radiology report quality control and provides a reproducible, verifiable methodological foundation for automated clinical text quality assurance.

0 citationsRead paper