# James Weatherhead — Research & Publications

**Peer-reviewed publications · privacy-preserving clinical AI**

My research asks how large language models can be used on real patient data *without leaking a patient's identity* — building the benchmarks, privacy measures, and on-device tooling that let clinicians use modern AI without sending protected health information outside a HIPAA-secure environment. Three peer-reviewed papers (2026) below, each with a plain-language summary and BibTeX; official abstracts live at every DOI.

## Peer-reviewed publications

### 1. K-anonymity decay in multi-turn clinical large language model conversations
- **Authors:** **Weatherhead J**, Hasan A, Weatherhead J, Golovko G, Grant B, Garcia JD, Certuche HS, Powell RP, Abril JM, McCaffrey P. *(first author)*
- **Venue:** *Frontiers in Digital Health.* 2026;8:1832168.
- **DOI:** https://doi.org/10.3389/fdgth.2026.1832168
- **Summary:** Measures how re-identification risk changes as a clinical conversation unfolds: individually harmless turns accumulate into identifying combinations, and the paper quantifies how quickly k-anonymity decays across multi-turn LLM dialogue — motivating de-identification that accounts for context, not just single messages.

```bibtex
@article{weatherhead2026kanonymity,
  author  = {Weatherhead, James and Hasan, A. and Weatherhead, J. and Golovko, G.
             and Grant, B. and Garcia, J. D. and Certuche, H. S. and Powell, R. P.
             and Abril, J. M. and McCaffrey, P.},
  title   = {K-anonymity decay in multi-turn clinical large language model conversations},
  journal = {Frontiers in Digital Health},
  year    = {2026},
  volume  = {8},
  pages   = {1832168},
  doi     = {10.3389/fdgth.2026.1832168}
}
```

### 2. ASQ-PHI: An adversarial synthetic data benchmark for clinical de-identification and search utility
- **Authors:** **Weatherhead J**, Golovko G, McCaffrey P. *(first author)*
- **Venue:** *Data in Brief.* 2026;65:112586.
- **DOI:** https://doi.org/10.1016/j.dib.2026.112586
- **Summary:** An adversarial, fully synthetic benchmark for clinical PHI de-identification. It stress-tests rule-based, cloud, vector, and on-device (Ollama) redaction and pairs de-identification accuracy with downstream search utility — so methods can be compared, and released as open data, without ever exposing a real patient record.

```bibtex
@article{weatherhead2026asqphi,
  author  = {Weatherhead, James and Golovko, G. and McCaffrey, P.},
  title   = {{ASQ-PHI}: An adversarial synthetic data benchmark for clinical
             de-identification and search utility},
  journal = {Data in Brief},
  year    = {2026},
  volume  = {65},
  pages   = {112586},
  doi     = {10.1016/j.dib.2026.112586}
}
```

### 3. Evaluation of multiple generative large language models on neurology board-style questions
- **Authors:** Almomani M, Valaparla V, **Weatherhead J**, Fang X, Dabi A, Li C-Y, McCaffrey P, Hier D, Rodríguez-Fernández JM. *(co-author)*
- **Venue:** *Frontiers in Digital Health.* 2026;7:1737882.
- **DOI:** https://doi.org/10.3389/fdgth.2025.1737882
- **Summary:** Benchmarks several generative LLMs on neurology board-style questions, comparing accuracy across models on specialty-exam content to gauge where general-purpose models stand on high-stakes clinical reasoning.

```bibtex
@article{almomani2026neurology,
  author  = {Almomani, M. and Valaparla, V. and Weatherhead, James and Fang, X.
             and Dabi, A. and Li, C.-Y. and McCaffrey, P. and Hier, D.
             and Rodr{\'i}guez-Fern{\'a}ndez, J. M.},
  title   = {Evaluation of multiple generative large language models on
             neurology board-style questions},
  journal = {Frontiers in Digital Health},
  year    = {2026},
  volume  = {7},
  pages   = {1737882},
  doi     = {10.3389/fdgth.2025.1737882}
}
```

## Conference presentations
- **AMIA 2026 Amplify, Denver — May 2026 (oral, Session CI25):** Local PHI Scrubber: On-Device PHI Redaction in Clinical LLM Workflows.
- **AMIA 2026 Amplify, Denver — May 2026 (poster):** ASQ-PHI: An Adversarial Synthetic Benchmark for Clinical PHI De-Identification (sponsored by the CU Anschutz Center for Health AI).
- **AI in Health Conference (AIHC), Rice University — Sep 2026 (poster):** Agentic AI-Assisted Real-World Evidence: EBV-Specific Neurovascular Risk after Corticosteroid Exposure in Older Adults (Golovko G, Cwiklik E, Watson R, Weatherhead J, Naydanova E, Botnar K, Woolsey C, Khanipov K).
- **Digital Pathology Association Conference — Oct 2025:** Beyond Digital Pathology: AI for Cancer Detection and LLMs to Personalize Care.

## Links
- Full profile: https://james.jcweatherhead.com/
- Full CV: https://james.jcweatherhead.com/resume/ (Markdown: https://james.jcweatherhead.com/resume/index.md)
- Structured profile (schema.org): https://james.jcweatherhead.com/profile.json
- Curated agent index: https://james.jcweatherhead.com/llms.txt