mlsanityData to 7 Oct 2026

Harmonized Feature Conditioning and Frequency-Prompt Personalization for Multi-Rater Medical Segmentation

arXiv 2605.08210 ↗First posted 6 May 2026cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

Multi-rater medical image segmentation captures the inherent ambiguity of clinical interpretation, where diagnostic boundaries vary across experts and imaging devices. Existing approaches often reduce this diversity to consensus labels or treat rater differences as noise, resulting in overconfident and poorly calibrated models. We propose a harmonized probabilistic framework that disentangles acquisition artifacts from genuine annotator variability through adaptive feature conditioning and frequency-domain personalization. A lightweight Harmonizer Network implicitly models scanner-specific artifacts and performs dynamic feature modulation to standardize latent representations, ensuring that uncertainty reflects anatomy rather than noise. To represent rater-specific styles, we introduce a novel High-Frequency Prompt Modules that operate in the spectral domain to encode annotator-dependent boundary precision and textural sensitivity. These prompts adaptively modulate harmonized features to produce personalized yet anatomically consistent segmentations. Furthermore, a Generalized Energy Distance based regularization aligns the generative distribution with empirical annotation variability, promoting diversity where experts disagree and consensus where they converge. Experiments on LIDC-IDRI and NPC-170 show SOTA aggregated and individualized segmentation, with notable GED reductions and improved Dice scores, especially on noisy cases. Beyond accuracy, the model exhibits clinically meaningful uncertainty. Confidence rises in agreement regions and declines in ambiguous areas, supporting its use as a reliable and interpretable tool for multi-expert clinical workflows.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 37 of its 45 references that are in the map. First posted on arXiv 6 May 2026.

Builds on

37 of 45 references.

References · 37

  1. 2017On Calibration of Modern Neural Networkscited by 4,860
  2. 2018Co-teaching: Robust Training of Deep Neural Networks with Extremely Noisy Labelscited by 1,053
  3. 2019Kvasir-SEG: A Segmented Polyp Datasetcited by 588
  4. 2019Symmetric Cross Entropy for Robust Learning with Noisy Labelscited by 499
  5. 2020Combating noisy labels by agreement: A joint training method with co-regularizationcited by 289
  6. 2018A Probabilistic U-Net for Segmentation of Ambiguous Imagescited by 281
  7. 2022Visual Attention Networkcited by 210
  8. 2018Not-so-supervised: a survey of semi-supervised, multi-instance, and transfer learning in medical image analysiscited by 178
  9. 2019PHiSeg: Capturing Uncertainty in Medical Image Segmentationcited by 128
  10. 2019Learning From Noisy Labels By Regularized Estimation Of Annotator Confusioncited by 126
  11. 2017The Cramer Distance as a Solution to Biased Wasserstein Gradientscited by 116
  12. 2020Stochastic Segmentation Networks: Modelling Spatially Correlated Aleatoric Uncertaintycited by 100
25 more
  1. 2023Ambiguous Medical Image Segmentation using Diffusion Modelscited by 92
  2. 2017Who Said What: Modeling Individual Labelers Improves Classificationcited by 77
  3. 2019A Hierarchical Probabilistic U-Net for Modeling Multi-Scale Ambiguitiescited by 65
  4. 2021Adaptive Early-Learning Correction for Segmentation from Noisy Annotationscited by 61
  5. 2020Disentangling Human Error from the Ground Truth in Segmentation of Medical Imagescited by 58
  6. 2021Is segmentation uncertainty useful?cited by 37
  7. 2021Superpixel-guided Iterative Learning from Noisy Labels for Medical Image Segmentationcited by 25
  8. 2021Using Soft Labels to Model Uncertainty in Medical Image Segmentationcited by 18
  9. 2023Probabilistic Modeling of Inter- and Intra-observer Variability in Medical Image Segmentationcited by 17
  10. 2020D-LEMA: Deep Learning Ensembles from Multiple Annotations -- Application to Skin Lesion Segmentationcited by 16
  11. 2023That Label's Got Style: Handling Label Style Bias for Uncertain Image Segmentationcited by 13
  12. 2024QUBIQ: Uncertainty Quantification for Biomedical Image Segmentation Challengecited by 12
  13. 2023Laplacian-Former: Overcoming the Limitations of Vision Transformers in Local Texture Detectioncited by 10
  14. 2024Diversified and Personalized Multi-rater Medical Image Segmentationcited by 9
  15. 2024Ultrasound Nodule Segmentation Using Asymmetric Learning with Simple Clinical Annotationcited by 9
  16. 2023Transformer-based Annotation Bias-aware Medical Image Segmentationcited by 8
  17. 2023Learning Robust Medical Image Segmentation from Multi-source Annotationscited by 4
  18. 2023Confidence Contours: Uncertainty-Aware Annotation for Medical Semantic Segmentationcited by 4
  19. 2022Calibrate the inter-observer segmentation uncertainty via diagnosis-first principlecited by 3
  20. 2023TAX: Tendency-and-Assignment Explainer for Semantic Segmentation with Multi-Annotatorscited by 3
  21. 2023Inter-Rater Uncertainty Quantification in Medical Image Segmentation via Rater-Specific Bayesian Neural Networkscited by 3
  22. 2024Multi-rater Prompting for Ambiguous Medical Image Segmentationcited by 3
  23. 2023Investigating and Improving Latent Density Segmentation Models for Aleatoric Uncertainty Quantification in Medical Imagingcited by 2
  24. 2025DiffOSeg: Omni Medical Image Segmentation via Multi-Expert Collaboration Diffusion Modelcited by 2
  25. 2025From Noisy Labels to Intrinsic Structure: A Geometric-Structural Dual-Guided Framework for Noise-Robust Medical Image Segmentationcited by 1

Built on by

Not cited yet.