Trusted Multi-Rater Segmentation

Sona Bezirganyan, Varduhi Yeghiazaryan

Keywords: uncertainty quantification, evidential deep learning, medical image segmentation, multi-rater segmentation

Abstract: Deep learning models have shown strong performance in medical image segmentation. But their integration into clinical practice has been slow, largely due to the lack of reliable uncertainty estimates. In medical imaging, uncertainty arises not only from the input data, but also from inter-rater variability in annotations. Most existing multi-rater segmentation approaches focus on modeling label disagreement through probabilistic outputs, without providing explicit uncertainty estimates. We propose Trusted Multi-Rater Segmentation (TMS), a novel algorithm that integrates evidential deep learning into multi-rater medical image segmentation. TMS treats network outputs associated with each annotator as subjective opinions, represented as parameters of the Dirichlet distribution, and combines them using weighted belief fusion from subjective logic. Unlike prior methods, TMS produces both probabilistic segmentations and explicit, interpretable uncertainty estimates. We demonstrate state-of-the-art performance in the optic disc and cup segmentation tasks using the RIGA dataset, as well as lung nodule segmentation using the LIDC dataset. Moreover, we go beyond conventional performance measures by explicitly evaluating the quality of uncertainty estimates, showing that TMS exhibits strong uncertainty-awareness.

Primary Area: probabilistic methods (Bayesian methods, variational inference, sampling, UQ, etc.)

Submission Number: 3802