Relation U-Net

2025-01-15

AI Summary

In medical image segmentation, model confidence often misaligns with actual accuracy, hindering clinically trustworthy deployment. To address this, we propose Relation U-Net—the first method to leverage pairwise structural relationships among multiple segmentation outputs for label-free confidence estimation. Built upon the U-Net architecture, it incorporates a novel relation encoding module that explicitly models both structural consistency and divergence across ensemble predictions, enabling end-to-end confidence prediction without ground-truth labels at test time. Our approach achieves strong linear correlation (r > 0.9) between predicted confidence and segmentation accuracy. Evaluated on four public medical imaging benchmarks, Relation U-Net surpasses standard U-Net in segmentation performance while delivering well-calibrated and interpretable uncertainty quantification. This establishes a new paradigm for safe, reliable AI-assisted diagnosis grounded in principled uncertainty modeling.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Relational Probabilistic Models
• Computer Vision: Segmentation

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Towards clinical interpretations, this paper presents a new ''output-with-confidence'' segmentation neural network with multiple input images and multiple output segmentation maps and their pairwise relations. A confidence score of the test image without ground-truth can be estimated from the difference among the estimated relation maps. We evaluate the method based on the widely used vanilla U-Net for segmentation and our new model is named Relation U-Net which can output segmentation maps of the input images as well as an estimated confidence score of the test image without ground-truth. Experimental results on four public datasets show that Relation U-Net can not only provide better accuracy than vanilla U-Net but also estimate a confidence score which is linearly correlated to the segmentation accuracy on test images.

Problem

• Image Understanding
• Model Confidence
• Medical Image Analysis

Innovation

• Relation U-Net
• Multi-image Processing
• Confidence Estimation