Using Artificial Intelligence to Label Free-Text Operative and Ultrasound Reports for Grading Pediatric Appendicitis
Bibliographic record
Abstract
PURPOSE: Data science approaches personalizing pediatric appendicitis management are hampered by small datasets and unstructured electronic medical records (EMR). Artificial intelligence (AI) chatbots based on large language models can structure free-text EMR data. We compare data extraction quality between ChatGPT-4 and human data collectors. METHODS: To train AI models to grade pediatric appendicitis preoperatively, several data collectors extracted detailed preoperative and operative data from 2100 children operated for acute appendicitis. Collectors were trained for the task based on satisfactory Kappa scores. ChatGPT-4 was prompted to structure free text from 103 random anonymized ultrasound and operative records in the dataset using the set variables and coding options, and to estimate appendicitis severity grade from the operative report. A pediatric surgeon then adjudicated all data, identifying errors in each method. RESULTS: Within the 44 ultrasound (42.7%) and 32 operative reports (31.1%) discordant in at least one field, 98% of the errors were found in the manual data extraction. The appendicitis grade was erroneously assigned manually in 29 patients (28.2%), and by ChatGPT-4 in 3 (2.9%). Across datasets, the use of the AI chatbot was able to avoid misclassification in 59.2% of the records including both reports and extracted data approximately 40 times faster. CONCLUSION: AI chatbot significantly outperformed manual data extraction in accuracy for ultrasound and operative reports, and correctly assigned the appendicitis grade. While wider validation is required and data safety concerns must be addressed, these AI tools show significant promise in improving the accuracy and efficiency of research data collection. LEVELS OF EVIDENCE: Level III.
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 | 0.018 | 0.093 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.005 | 0.002 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.003 |
| Open science | 0.002 | 0.003 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 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".