Loading
Life sciences · Preprint
arXiv · September 9, 2026
Early or partial results. Treat as a signal, not a conclusion.
This preprint proposes Referee-Based Quality Estimation (RBQE), a deployment-time reliability signal for polyp segmentation that measures agreement between a primary model and an independently trained referee model. Evaluated on 1,223 external images, SegFormer-B0 cross-architecture referees achieved ROC-AUC 0.960 for detecting segmentation failures, outperforming same-architecture controls and test-time augmentation baselines. The approach is reference-free, requires only one additional forward pass, and shows improved retention of high-confidence predictions, but remains unvalidated in live clinical deployment and has not undergone peer review.
Algorithm validation study using external benchmark and referee configurations across two design axes. Images from four public polyp segmentation datasets; no clinical patient population or in vivo deployment context reported.. Intervention: Referee-Based Quality Estimation (RBQE) using cross-model agreement (primary model + independently trained referee) as reliability signal; Agreement Dice descriptor; selective prediction with progressive rejection of low-agreement cases.. Compared with: Same-architecture referee (random initialization only), UNet++ cross-architecture referee, MedSAM prompt-coupled referee, and Test-Time Augmentation baseline.. n = 1,223. Multi-source public datasets; specific geographic origins and clinical centres not stated..
Same-architecture referee (different random initialization only) achieved ROC-AUC = 0.923 for detecting model failure, demonstrating independent training alone provides useful signal SegFormer-B0 cross-architecture referee achieved strongest performance: ROC-AUC = 0.960, significantly outperforming same-architecture control and UNet++ SegFormer-B0 exceeded Test-Time Augmentation baseline by 0.055 ROC-AUC under identical protocol
Safety was not reported in the material analysed. Check the source before drawing any conclusion about harm.
If validated in prospective clinical deployment, RBQE could enable real-time detection of segmentation failures during live colonoscopy, reducing silent model errors and supporting selective prediction. However, the framework remains unproven in actual clinical workflow and requires peer-reviewed validation before clinical adoption.
A novel reference-free reliability framework for polyp segmentation evaluated on external benchmark data with promising ROC-AUC results, but lacking clinical validation, prospective deployment data, or peer review.
As stated by the source record.
Quoted from the source exactly as published.
If validated in prospective clinical deployment, RBQE could enable real-time detection of segmentation failures during live colonoscopy, reducing silent model errors and supporting selective prediction. However, the framework remains unproven in actual clinical workflow and requires peer-reviewed validation before clinical adoption.
Graded across the dimensions that decide whether you should act, each from what the source actually supports. There is no single score, and where a dimension was not assessed it says so.
In real-time colonoscopy, ground-truth annotations are unavailable at inference, so polyp segmentation models can fail silently. We propose Referee-Based Quality Estimation (RBQE), a reference-free framework measuring agreement between a primary segmentation model and an independently trained referee on the same image. RBQE is evaluated on a standardized 1,223-image external benchmark drawn from four public datasets, using four referee configurations chosen to separate two design axes: referee independence and architectural diversity. Using a common Agreement Dice descriptor, a same-architecture referee differing from the primary model only in random initialization already yields a useful reliability signal (ROC-AUC = 0.923), showing that independent training alone is sufficient. Cross-architecture referees improve further: SegFormer-B0 achieves the strongest performance (ROC-AUC = 0.960), significantly outperforming the same-architecture control and UNet++, and exceeding a representative Test-Time Augmentation baseline by 0.055 ROC-AUC under an identical protocol, whereas a prompt-coupled MedSAM referee underperforms despite maximal architectural diversity. Because empty-mask agreement is trivially separable, we also report a restricted evaluation excluding such cases: ROC-AUC falls to 0.876 (SegFormer-B0, 1,046 images) and 0.783 (same-architecture control, 975 images), yet RBQE's margin over both baselines widens on this identical subset. RBQE additionally increases the mean Dice of retained predictions as low-agreement cases are progressively rejected, supporting selective prediction, and requires only one additional deterministic referee forward pass at inference. Our study therefore supports cross-model agreement as a practical, interpretable reliability framework for automated polyp segmentation.
Taken from the source record, never inferred. Follow any of these and new work involving them reaches your briefing.