Performance of natural language processing in identifying adenomas from colonoscopy reports: a systematic review and meta-analysis
Bibliographic record
Abstract
Background and Aims: The adenoma detection rate is a key quality metric for colonoscopy and is inversely related to the post-colonoscopy colorectal cancer rate. Natural language processing can be used to automate the generation of such quality metrics from colonoscopy reports. We performed a systematic review and meta-analysis on the performance of natural language processing (NLP) in identifying adenoma detection in colonoscopy and paired pathology reports. Methods: We performed a systematic review and meta-analysis according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses recommendations. A literature query was conducted on MEDLINE, Embase, and Cochrane Database of Systematic Reviews through July 2022. Studies were included if they reported on the operator characteristics of an NLP algorithm in interpreting adenoma detection in colonoscopy and pathology reports. Two authors independently screened studies and abstracted data using an a priori designed data collection form. Performance characteristics were pooled by first using a univariate analysis, followed by a bivariate analysis of sensitivity and specificity. Results: The pooled specificity and sensitivity for identifying adenoma detection were .997 (95% confidence interval [CI], .984-.999) and .978 (95% CI, .938-.992). The pooled positive predictive value, negative predictive value, and F1 score were .997 (95% CI, .979-1.00), .977 (95% CI, .938-.992), and .982 (95% CI, .957-.993), respectively. In the bivariate analysis, the pooled specificity and sensitivity were .992 (95% CI, .978-.997) and .973 (95% CI, .929-.990). The NLP systems performed similarly well in identifying the detection of sessile serrated lesions and advanced adenomas. Conclusions: NLP systems can identify adenoma detection from colonoscopy and pathology reports with strong operator characteristics.
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 | 0.058 | 0.144 |
| Meta-epidemiology (narrow) | 0.003 | 0.002 |
| Meta-epidemiology (broad) | 0.019 | 0.059 |
| Bibliometrics | 0.010 | 0.008 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.005 | 0.004 |
| Open science | 0.003 | 0.002 |
| Research integrity | 0.003 | 0.002 |
| Insufficient payload (model declined to judge) | 0.003 | 0.000 |
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".