# Marginal Thresholding in Noisy Image Segmentation

arXiv: https://arxiv.org/abs/2304.04116 · HTML: https://mlsanity.com/paper/2304.04116/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 20 of its 35 references that are in the map.
- First posted on arXiv 8 Apr 2023.

## Builds on (most cited first, 20 of 20)

- [U-Net: Convolutional Networks for Biomedical Image Segmentation](https://mlsanity.com/paper/1505.04597/) (2015)
- [Embracing Imperfect Datasets: A Review of Deep Learning Solutions for Medical Image Segmentation](https://mlsanity.com/paper/1908.10454/) (2019)
- [Deep learning with noisy labels: exploring techniques and remedies in medical image analysis](https://mlsanity.com/paper/1912.02911/) (2019)
- [U-Net and its variants for medical image segmentation: theory and applications](https://mlsanity.com/paper/2011.01118/) (2020)
- [Confidence Calibration and Predictive Uncertainty Estimation for Deep Medical Image Segmentation](https://mlsanity.com/paper/1911.13273/) (2019)
- [Optimizing the Dice Score and Jaccard Index for Medical Image Segmentation: Theory & Practice](https://mlsanity.com/paper/1911.01685/) (2019)
- [SoftSeg: Advantages of soft versus binary training for image segmentation](https://mlsanity.com/paper/2011.09041/) (2020)
- [Learning to Segment Skin Lesions from Noisy Annotations](https://mlsanity.com/paper/1906.03815/) (2019)
- [Post Training Uncertainty Calibration of Deep Networks For Medical Image Segmentation](https://mlsanity.com/paper/2010.14290/) (2020)
- [Using Soft Labels to Model Uncertainty in Medical Image Segmentation](https://mlsanity.com/paper/2109.12622/) (2021)
- [Imperfect Segmentation Labels: How Much Do They Matter?](https://mlsanity.com/paper/1806.04618/) (2018)
- [Label fusion and training methods for reliable representation of inter-rater uncertainty](https://mlsanity.com/paper/2202.07550/) (2022)
- [Soft labeling by Distilling Anatomical knowledge for Improved MS Lesion Segmentation](https://mlsanity.com/paper/1901.09263/) (2019)
- [On the relationship between calibrated predictors and unbiased volume estimation](https://mlsanity.com/paper/2112.12560/) (2021)
- [Label noise in segmentation networks : mitigation must deal with bias](https://mlsanity.com/paper/2107.02189/) (2021)
- [Robustness study of noisy annotation in deep learning based medical image segmentation](https://mlsanity.com/paper/2003.06240/) (2020)
- [Superpixel-Guided Label Softening for Medical Image Segmentation](https://mlsanity.com/paper/2007.08897/) (2020)
- [On Image Segmentation With Noisy Labels: Characterization and Volume Properties of the Optimal Solutions to Accuracy and Dice](https://mlsanity.com/paper/2206.06484/) (2022)
- [Theoretical analysis and experimental validation of volume bias of soft Dice optimized segmentation maps in the context of inherent uncertainty](https://mlsanity.com/paper/2211.04161/) (2022)
- [Noisy Image Segmentation With Soft-Dice](https://mlsanity.com/paper/2304.00801/) (2023)

## Abstract (arXiv)

This work presents a study on label noise in medical image segmentation by considering a noise model based on Gaussian field deformations. Such noise is of interest because it yields realistic looking segmentations and because it is unbiased in the sense that the expected deformation is the identity mapping. Efficient methods for sampling and closed form solutions for the marginal probabilities are provided. Moreover, theoretically optimal solutions to the loss functions cross-entropy and soft-Dice are studied and it is shown how they diverge as the level of noise increases. Based on recent work on loss function characterization, it is shown that optimal solutions to soft-Dice can be recovered by thresholding solutions to cross-entropy with a particular a priori unknown threshold that efficiently can be computed. This raises the question whether the decrease in performance seen when using cross-entropy as compared to soft-Dice is caused by using the wrong threshold. The hypothesis is validated in 5-fold studies on three organ segmentation problems from the TotalSegmentor data set, using 4 different strengths of noise. The results show that changing the threshold leads the performance of cross-entropy to go from systematically worse than soft-Dice to similar or better results than soft-Dice.