Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-2210
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-10-23T06:46:24.536246Z
status: published
lens: g_space
sector: health
headline: Impression-regularized finetuning enables small language models to extract cervical spine fractures from radiology reports
dek: We assessed whether small, open-weight large language models (LLMs) can accurately extract vertebral-level acute cervical spine fracture labels from CT radiology reports while supporting on-premises deployment, and whether impression-regularized finetuning improves accuracy, data efficiency, and temporal stability. In this single-center retrospective study, 8,567 adult cervical spine CT reports (2017-2022) were labeled for acute occipital condyle and C1-C7 fractures with neuroradiologist oversight. Reports from…
gain_title: A regularized finetuning method raised small open-weight models' accuracy at extracting vertebral-level fracture labels, bringing a 4B model close to a much larger zero-shot model while allowing privacy-preserving local use.
problem_title: (none)
trace_subject: (none)
gain_reading: A regularized finetuning method raised small open-weight models' accuracy at extracting vertebral-level fracture labels, bringing a 4B model close to a much larger zero-shot model while allowing privacy-preserving local use.
gain_evidence: impression-regularized Gemma-3-4B approached the zero-shot performance of the Gemma-3-27B (macro-F1 0.909 vs. 0.913) | suitable for privacy-preserving on-premises deployment
problem_reading: (none)
problem_evidence: (none)
quick_read: Researchers tested whether small, open-weight language models could pull vertebral-level acute cervical spine fracture labels out of CT radiology reports. They compared standard low-rank adaptation with a version that adds an auxiliary loss based on synthetic, fracture-focused impressions, using 8,567 reports from one center collected between 2017 and 2022.

The approach matters because it suggests that modest models can run inside a hospital's own infrastructure rather than depending on large external systems, which supports privacy and auditability. The evidence comes from a single center and retrospective data, so performance elsewhere and in live clinical workflows is not established by this work.
limitation: The study is single-center and retrospective, and the remaining errors were mainly tied to ambiguous report language and label-boundary cases, so generalizability beyond this setting is not shown.
tag: Evidence-backed gain
key_points: Gemma-3-4B macro-F1 rose from 0.885 to 0.909 with the regularized method, and run-to-run variance fell. | The smallest model, Gemma-3-1B, improved from 0.646 to 0.782, showing the largest gains in small models and low-data settings. | On 2018-2022 temporal testing, mean F1 increased from 0.871 to 0.893 with similar or lower variability across time windows. | Remaining errors were mainly linked to ambiguous report language and label-boundary cases.
rundown: The dataset consisted of 8,567 adult cervical spine CT reports labeled for acute occipital condyle and C1-C7 fractures with neuroradiologist oversight. Only the 1,064 reports from 2017 were used for finetuning, while 7,503 reports from 2018-2022 served as a temporal test set.

Models tested included Gemma-3 (1B, 4B, 12B), Llama-3.2-3B, and Phi-3.5-Mini. During training the model samples several synthetic targeted impressions, picks the one most consistent with reference labels, and applies an auxiliary impression loss; at inference, each model does single-pass deterministic JSON extraction.

Baselines were zero-shot prompting, two-pass impression-guided prompting, and standard LoRA. The tuned 4B model reached macro-F1 of 0.909 against 0.913 for zero-shot Gemma-3-27B, with macro-precision 0.929 and macro-recall 0.893.
sources:
- peer_reviewed | PLOS Digital Health | https://doi.org/10.1371/journal.pdig.0001786 | 2026-10-21
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

836456391deed8f5f7ec0780605d214e4518e43b81cc6f78a79a639597f0d8e0

previous

0000000000000000000000000000000000000000000000000000000000000000