Decoupled Single-Mask Annotation Noise Detection via Cross-Sectional Patch Self-Consistency
For medical imaging researchers and practitioners, the method provides an interpretable, auditable tool to identify annotation noise in single-mask datasets, addressing a pervasive but under-addressed problem.
The paper tackles single-mask annotation noise in vascular CT datasets, introducing a decoupled framework that uses cross-sectional patch self-consistency to detect and localize noisy labels without requiring multirater fusion or network coupling. Experiments on coronary CT data show the method improves training robustness and reveals systematic biases, e.g., transverse/oblique vessels have 5.1× higher error rates.
Vascular computed tomography datasets are commonly annotated only once per scan, yielding the pervasive yet under addressed problem of single mask annotation noise. Existing solutions either require costly multirater fusion or are coupled with network training, preventing explicit auditing of where and why labels fail. We introduce a decoupled framework for single-mask annotation noise detection that leverages cross-sectional patch self-consistency to produce interpretable and auditable noise evidence. Tubular anatomy exhibits strong cross-sectional recurrence: patches extracted orthogonally along vessel centrelines recur in appearance across locations and subjects. Thus, anatomically similar patches should have consistent masks, and disagreement signals unreliable annotation. Our method samples cross-sectional patches, retrieves intensity-equivalent neighbours via scalable vector search, and computes a patch-level noise score from statistical mask disagreement, yielding explicit image-mask evidence for every flagged region. Aggregating scores produces scan-level quality maps for dataset quality assessment or quality-weighted training. Experiments on the coronary CT dataset validate the detected noise for improving training robustness and reveal systematic annotation biases. Specifically, transverse and oblique vessels exhibit 5.1 times higher error rates than axis-aligned structures, with additional correlations to cross-sectional area and intensity. Code is available here.