Adapting Large Language Models for Automatic Annotation of Radiology Reports for Metastases Detection
Bibliographic record
Abstract
Automatic identification of metastatic sites in cancer patients from electronic health records is a challenging yet crucial task with significant implications for diagnosis and treatment. In this study, we propose a method to detect metastases from non-structured radiology report texts by accessing only their impression section. We build models based on pre-trained large language models and parameter-efficient fine-tuning. We compare model performances between utilizing non-structured reports and reports following institutional-level templates. By incorporating patient historical data and their timeline into the model, we bridge the gap between structured and non-structured reports. Our experiments are conducted on data gathered at Memorial Sloan Kettering Cancer Center (MSKCC) which have been annotated for metastases presence in three organs: liver, lung, and adrenal glands. Our results suggest that access to previous reports significantly improves model performance, with an average improvement of 7.7 points in terms of F1-score over all datasets. Additionally, incorporating temporal information enhances the accuracy of metastasis detection by 0.4 and 1.1 points on liver and adrenal glands data, respectively. Our method shows potential for automating radiology report labeling on a large scale in an efficient manner, with the potential to deploy on low-cost hardware.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.010 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.003 | 0.002 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.001 | 0.003 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".