Research

Paper

TESTING March 12, 2026

Reproducible Synthetic Clinical Letters for Seizure Frequency Information Extraction

Authors

Yujian Gan, Stephen H. Barlow, Ben Holgate, Joe Davies, James T. Teo, Joel S. Winston, Mark P. Richardson

Abstract

Seizure-frequency information is important for epilepsy research and clinical care, but it is usually recorded in variable free-text clinic letters that are hard to annotate and share. We developed a reproducible, privacy-preserving framework for extracting seizure frequency using fully synthetic yet task-faithful epilepsy letters. We defined a structured label scheme covering common descriptions of seizure burden, including explicit rates, ranges, clusters, seizure-free intervals, unknown frequency, and explicit no-seizure statements. A teacher language model generated NHS-style synthetic letters paired with normalized labels, rationales, and evidence spans. We fine-tuned several open-weight language models (4B-14B parameters) on these synthetic letters to extract seizure frequency from full documents, comparing direct numeric prediction with structured label prediction and testing evidence-grounded outputs. On a clinician-checked held-out set of real clinic letters, models trained only on synthetic data generalized well, and structured labels consistently outperformed direct numeric regression. With 15,000 synthetic training letters, models achieved micro-F1 scores up to 0.788 for fine-grained categories and 0.847 for pragmatic categories; a medically oriented 4B model achieved 0.787 and 0.858, respectively. Evidence-grounded outputs also supported rapid clinical verification and error analysis. These results show that synthetic, structured, evidence-grounded supervision can enable robust seizure-frequency extraction without sharing sensitive patient text and may generalize to other temporally complex clinical information extraction tasks.

Metadata

arXiv ID: 2603.11407
Provider: ARXIV
Primary Category: cs.IR
Published: 2026-03-12
Fetched: 2026-03-13 06:02

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.11407v1</id>\n    <title>Reproducible Synthetic Clinical Letters for Seizure Frequency Information Extraction</title>\n    <updated>2026-03-12T00:41:21Z</updated>\n    <link href='https://arxiv.org/abs/2603.11407v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.11407v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>Seizure-frequency information is important for epilepsy research and clinical care, but it is usually recorded in variable free-text clinic letters that are hard to annotate and share. We developed a reproducible, privacy-preserving framework for extracting seizure frequency using fully synthetic yet task-faithful epilepsy letters. We defined a structured label scheme covering common descriptions of seizure burden, including explicit rates, ranges, clusters, seizure-free intervals, unknown frequency, and explicit no-seizure statements. A teacher language model generated NHS-style synthetic letters paired with normalized labels, rationales, and evidence spans. We fine-tuned several open-weight language models (4B-14B parameters) on these synthetic letters to extract seizure frequency from full documents, comparing direct numeric prediction with structured label prediction and testing evidence-grounded outputs. On a clinician-checked held-out set of real clinic letters, models trained only on synthetic data generalized well, and structured labels consistently outperformed direct numeric regression. With 15,000 synthetic training letters, models achieved micro-F1 scores up to 0.788 for fine-grained categories and 0.847 for pragmatic categories; a medically oriented 4B model achieved 0.787 and 0.858, respectively. Evidence-grounded outputs also supported rapid clinical verification and error analysis. These results show that synthetic, structured, evidence-grounded supervision can enable robust seizure-frequency extraction without sharing sensitive patient text and may generalize to other temporally complex clinical information extraction tasks.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.IR'/>\n    <published>2026-03-12T00:41:21Z</published>\n    <arxiv:primary_category term='cs.IR'/>\n    <author>\n      <name>Yujian Gan</name>\n    </author>\n    <author>\n      <name>Stephen H. Barlow</name>\n    </author>\n    <author>\n      <name>Ben Holgate</name>\n    </author>\n    <author>\n      <name>Joe Davies</name>\n    </author>\n    <author>\n      <name>James T. Teo</name>\n    </author>\n    <author>\n      <name>Joel S. Winston</name>\n    </author>\n    <author>\n      <name>Mark P. Richardson</name>\n    </author>\n  </entry>"
}