Confidence-Based Annotation Of Brain Tumours In Ultrasound

2025-02-21

AI Summary

Discrete segmentation annotations for brain tumor ultrasound images suffer from high inter-observer variability and stochastic uncertainty—particularly at tumor boundaries—rendering them inherently unreliable. Method: This work theoretically establishes, for the first time, the intrinsic infeasibility of discrete annotation under B-mode ultrasound imaging conditions. We propose a sparse confidence-labeling paradigm that integrates radiological priors with computer vision principles, using pixel-wise confidence scores as soft labels to supervise deep segmentation networks. Furthermore, we design an uncertainty calibration evaluation framework grounded in the Brier score. Results: Experiments demonstrate a strong linear correlation (Pearson *r* = 0.8) between boundary confidence and inter-observer variance. Across all cross-validation folds, models trained on confidence labels achieve significantly lower Brier scores than those trained on hard labels, thereby enhancing segmentation robustness and uncertainty quantification capability.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Segmentation
• Reasoning under Uncertainty: Uncertainty Representations

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Security and Privacy: Large-scale security measurements

Abstract

Purpose: An investigation of the challenge of annotating discrete segmentations of brain tumours in ultrasound, with a focus on the issue of aleatoric uncertainty along the tumour margin, particularly for diffuse tumours. A segmentation protocol and method is proposed that incorporates this margin-related uncertainty while minimising the interobserver variance through reduced subjectivity, thereby diminishing annotator epistemic uncertainty. Approach: A sparse confidence method for annotation is proposed, based on a protocol designed using computer vision and radiology theory. Results: Output annotations using the proposed method are compared with the corresponding professional discrete annotation variance between the observers. A linear relationship was measured within the tumour margin region, with a Pearson correlation of 0.8. The downstream application was explored, comparing training using confidence annotations as soft labels with using the best discrete annotations as hard labels. In all evaluation folds, the Brier score was superior for the soft-label trained network. Conclusion: A formal framework was constructed to demonstrate the infeasibility of discrete annotation of brain tumours in B-mode ultrasound. Subsequently, a method for sparse confidence-based annotation is proposed and evaluated. Keywords: Brain tumours, ultrasound, confidence, annotation.

Problem

• Addressing uncertainty in brain tumour segmentation
• Reducing subjectivity in ultrasound annotations
• Enhancing accuracy with confidence-based annotation methods

Innovation

• Sparse confidence method
• Reduces interobserver variance
• Soft-label trained network