My research sits at the intersection of natural language processing, clinical informatics, and veterinary epidemiology. The thread connecting most of my work is a simple question: can we use the data already being generated in clinical practice to make medicine better?
Antimicrobial resistance is one of the most pressing global health threats. Companion animal veterinary practices generate millions of clinical records, but until our work, there was no way to analyze prescribing patterns at scale. Working with VetCompass Australia, we built NLP pipelines to extract antimicrobial usage from over 4.4 million consultation records across 180+ clinics.
This started with rule-based extraction methods (2019), scaled to population-level analysis across Australia (2020, PLOS ONE — our most-cited work), and deepened into specific drugs like cefovecin, a critically important third-generation cephalosporin. We then developed VetBERT, a domain-adapted language model that minimized annotation effort for classifying disease syndromes in veterinary text (BioNLP @ ACL 2020).
The clinical impact followed: we evaluated doses and guideline agreement across millions of records (JAC-AMR 2022), and our methods supported a 135-clinic antimicrobial stewardship trial — one of the largest veterinary AMS interventions ever conducted.
More recently, my work has expanded beyond veterinary medicine. At the University of Washington, I've been involved in speech-based dementia detection (ACL 2025), evaluating LLMs on veterinary prescription analysis (BioNLP @ ACL 2024), and addressing bias in foundation models trained on multi-institutional datasets. Our latest work proposes a probabilistic paradigm for navigating the evaluation gaps that arise when applying AI to medicine.
Anusha Elangovan, Lei Xu, Morteza Elyasi, Ilker Akdulum, Mehmet Aksakal, Engin Gurun, Brian Hur
Proposed a probabilistic paradigm for evaluating AI and LLMs in medicine, addressing challenges of elusive ground truth and uncertainty.
Zhonghao Sheng, Xin Ding, Brian Hur, Changye Li, Trevor Cohen, Serguei V.S. Pakhomov
Proposed weight masking techniques to mitigate confounding factors in speech-based models for dementia detection.
Xin Ding, Zhonghao Sheng, Brian Hur, Jason Tauscher, Dror Ben-Zeev, Meliha Yetisgen, Serguei V.S. Pakhomov, Trevor Cohen
Applied task arithmetic techniques to address bias arising from multi-institutional training data in biomedical foundation models.
Suzanna Richards, Kirsten E. Bailey, Riata Scarborough, James R. Gilkerson, Glenn F. Browning, Brian Hur, Laura Y. Hardefeldt
Cross-sectional evaluation of a large-scale antimicrobial stewardship intervention across Australian companion animal practices.
Brian Hur, Lucy Lu Wang, Laura Hardefeldt, Meliha Yetisgen-Yildiz
Evaluated the performance of large language models on analyzing veterinary prescription text, assessing whether LLMs can accurately extract dosing information.
Brian Hur, Karin M. Verspoor, Timothy Baldwin, Laura Y. Hardefeldt, Christine Pfeiffer, Caroline Mansfield, James R. Gilkerson
Combined NLP with patient journey clustering to identify temporal patterns in antimicrobial treatments for cat bite abscesses across veterinary clinics.
Xin Ding, Zhonghao Sheng, Brian Hur, Fengyi Chen, Serguei V.S. Pakhomov, Trevor Cohen
Investigated methods to enhance the robustness of foundation model representations under distribution shifts related to data provenance.
Laura Y. Hardefeldt, Brian Hur, Suzanna Richards, Riata Scarborough, Glenn F. Browning, Kirsten E. Bailey, James R. Gilkerson
Large-scale implementation trial of antimicrobial stewardship interventions across 135 veterinary clinics in Australia, evaluating the effectiveness of evidence-based guidelines.
Brian Hur, Laura Y. Hardefeldt, Karin M. Verspoor, Timothy Baldwin, James R. Gilkerson
Used NLP to extract antimicrobial doses and indications from millions of veterinary records, comparing actual prescribing patterns against clinical guidelines to inform stewardship programs.
Brian Hur, Laura Y. Hardefeldt, Karin Verspoor, Timothy Baldwin, James R. Gilkerson
Addressed the technical challenges of applying deep learning to extract prescribing information from large-scale, noisy veterinary clinical text data.
Karin A. Thursky, Laura Y. Hardefeldt, Aparna Rajkhowa, Clare Ierano, Jonathan Bishop, Lesley Hawes, Brian Hur
Explored the role of qualitative research in developing antimicrobial stewardship programs across human and veterinary medicine in Australia.
Yuxia Wang, Brian Hur, Karin Verspoor, Timothy Baldwin
Developed the highest-performing rules-based method for clinical concept normalization in the N2C2 shared task, contrasting it with ClinicalBERT-based approaches.
Brian A. Hur, Laura Y. Hardefeldt, Karin M. Verspoor, Timothy Baldwin, James R. Gilkerson
Developed NLP methods to extract and describe antimicrobial usage patterns from 4.4 million veterinary consultation records across 180+ Australian clinics, establishing the foundation for large-scale antimicrobial stewardship research.
Brian Hur, Timothy Baldwin, Karin Verspoor, Laura Hardefeldt, James Gilkerson
Proposed instance selection and VetBERT, a domain-adapted language model, to minimize annotation effort for classifying the reason for antimicrobial administration in veterinary clinical notes.
Laura Y. Hardefeldt, Brian Hur, Karin Verspoor, Timothy Baldwin, Kirsten E. Bailey, Riata Scarborough, Suzanna Richards, Helen Billman-Jacobe, Glenn F. Browning, James R. Gilkerson
Described the usage patterns of cefovecin, a critically important third-generation cephalosporin, in Australian veterinary practices using NLP-generated labels from clinical records.
Brian Hur, Yuxia Wang, Timothy Baldwin, Karin Verspoor
Compared traditional n-gram matching approaches with ClinicalBERT for the task of normalizing medical concepts to standardized terminologies.
Brian Hur, Laura Y. Hardefeldt, Karin Verspoor, Timothy Baldwin, James R. Gilkerson
Preliminary study developing and testing machine learning and rules-based methods for extracting antimicrobials from clinical records stored within VetCompass across 180+ Australian clinics.
Antimicrobial prescribing in dogs and cats with urinary tract disease in a prospective intervention trial
Journal of Veterinary Internal Medicine, 2026
Retrospective cohort study on the development of keratoconjunctivitis sicca in dogs treated with trimethoprim sulfonamide: a VetCompass Australia study
Journal of Veterinary Internal Medicine, 2026