Pre-operative radiomics model for prognostication in resectable pancreatic adenocarcinoma: multi-institutional development and external
Bibliographic record
Abstract
Objectives: In patients with resectable pancreatic ductal adenocarcinoma (PDAC), there are few strictly pre-operative prognostic biomarkers available to guide therapy decisions. Radiomics has demonstrated potential prognostic value but it lacks external validation. We aimed to develop and externally validate a pre-operative clinical-radiomic prognostic model for PDAC. Methods: This was a retrospective international, multi-center study in patients with resectable PDAC who underwent pre-operative contrast-enhanced CT. Patients who received neoadjuvant therapy were excluded. The training cohort consisted of 352 patients who underwent CTs at five Toronto hospitals and subsequent resection at Toronto General Hospital, Toronto, Canada. The external test cohort consistent of 215 patients who underwent resection at a St Vincent’s University Hospital, Dublin, following pre-operative CTs performed at 34 Irish hospitals. Segmentation was performed using 3d Slicer v 4.11.2. Then 116 radiomic features were extracted using the PyRadiomics 3.0 library. Pre-operative Cox proportional hazard models incorporated (a) clinical factors (clinical), (b) clinical plus radiomics features (clinical-radiomic) and (c) a post-operative model incorporating pathological findings (TNM), which served as the reference standard. Outcomes were overall (OS) and disease-free survival (DFS). Model discrimination and calibration were assessed using concordance index (C-index), calibration plots and mean calibration error. A previously validated statistical tool for batch-effect correction (Combat) was used in an attempt to mitigate the impact of variation in CT scanner protocols between the multiple study sites. Results: In the validation cohort, the Radiomic signature was predictive of OS / DFS, with adjusted hazard ratios (HR) of 2.87 (95% CI: 1.40-5.87, p<0.001 / 5.28 (95% CI 2.35-11.86, p<0.001) respectively, along with age 1.02 (1.01-1.04, p=0.01) / 1.02 (1.00-1.04, p=0.03). No other clinical features were significantly associated with OS and DFS. Median OS was 22.9 versus 37 months (p=0.0092) and DFS 14.2 versus 29.8 Abstract 7 (p=0.0023) for the high versus low-risk groups in the external cohort. Calibration was moderate in the external cohort, with mean absolute error 7% and 13% for OS at 3 and 5 years respectively. The clinical-radiomic model demonstrated better discrimination for OS in the external cohort (C-index 0.545, 95%: 0.543-0.546) than the clinical model alone (0.497 95% CI 0.496-0.499, p<0.001) or the post-operative TNM model (0.525 95% CI 0.534-0.526, p<0.001). Implementation of Combat to mitigate the impact of multi-institutional variation in CT acquisition parameters did not improve discrimination results. In decision curve analysis, despite superior net benefit compared to clinical model, the clinical-radiomic model was not clinically useful for most threshold probabilities. TNM demonstrated the highest net benefit of the three models. Conclusion: A pre-operative model containing clinical variables and radiomics significantly improved prognostication of patients with resectable PDAC compared to using clinical information alone and it generalized to a large external dataset. Performance was similar to using pathological data (TNM), which are only available post-operatively. Despite superior performance compared to the clinical model, discrimination and clinical utility are suboptimal. This likely reflects inherent limitations of radiomics for PDAC prognostication, when deployed in real-world settings. Future work should focus upon standardization of CT acquisition protocols.
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 distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.000 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.003 | 0.001 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 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 teacher head, 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".