Natural language processing for local, regional, and distant breast cancer relapse identification in pathology reports
Bibliographic record
Abstract
PURPOSE: Cancer registries rarely track breast cancer relapse due to the resource-intensive nature of manual chart review. To address this gap, we developed natural language processing (NLP) models to automate the identification of breast cancer relapse in pathology reports. METHODS: We collected pathology reports from patients diagnosed with breast cancer between January 1, 2005, and December 31, 2014, in British Columbia, Canada, and manually annotated each for the presence or absence of local, regional, distant, and any breast cancer relapses. With these reports, we fine-tuned large language models to classify pathology reports. RESULTS: The corpus contained 1,888 pathology reports from a cohort of 993 breast cancer patients. Of these reports, 673 (35.6%) described local, 296 (15.7%) regional, and 654 (34.6%) distant relapses. In addition, 1,510 (80.0%) described at least one of any relapse type. The median time from diagnosis to first relapse was 7.3 years (range 0.2-18.2). All models demonstrated excellent performance. The local-relapse model performed particularly well, with > 93% accuracy, sensitivity, specificity, and 0.98 area under the receiver operating characteristic curve (AUC) score. CONCLUSION: We developed NLP models to detect breast cancer relapses from pathology reports with excellent accuracy, sensitivity, specificity, and AUC. NLP may facilitate more efficient and accurate collection of breast cancer outcomes data from clinical reports.
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 | 0.012 | 0.048 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.008 | 0.003 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.003 | 0.003 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.001 | 0.001 |
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".