Beyond Boundary Noise: Aggregated Aleatoric Uncertainty Fails to Capture Presence Ambiguity in 3D Lung Nodule Segmentation

2026-08-14

AI Summary

This study addresses the limitation of entropy-based uncertainty in capturing lesion existence ambiguity during 3D pulmonary nodule segmentation. We reveal that conventional entropy aggregation primarily reflects boundary noise, whereas existence ambiguity is inherently encoded within encoder features. Accordingly, we propose a lightweight supervised ambiguity head that leverages frozen features to precisely identify and rectify theoretical cognitive biases. Integrated with MC Dropout and deep ensembles, our method significantly outperforms entropy aggregation baselines across multiple architectures and external validation sets, achieving performance comparable to explicit modeling approaches. This work establishes an efficient new paradigm for uncertainty quantification in medical imaging by effectively bridging the gap between feature-level ambiguity representation and reliable uncertainty estimation without imposing substantial computational overhead.

Abstract

Uncertainty estimation is critical for the safe clinical deployment of deep learning in medical image segmentation, with aleatoric uncertainty theoretically designed to capture irreducible data ambiguity. However, whether entropy-based measures reflect clinically meaningful ambiguity, i.e. case-level disagreement about whether a pathology is present at all, remains poorly understood. Contrary to most prior work, which focused on pixel-wise boundary disagreement, we systematically evaluate how well aleatoric uncertainty captures presence ambiguity. Our evaluation spans 3D lung nodule segmentation across four architectures with Monte Carlo dropout and deep ensembles, on LIDC-IDRI and an external validation cohort (LNDb). We find that entropy-based uncertainty maps align with boundary noise and minor drawing variation but carry insufficient discriminative signal for presence ambiguity. In contrast, a lightweight supervised ambiguity head trained on frozen segmentation features substantially outperforms all entropy-aggregation-based baselines across architectures, metrics, and both cohorts, and matches or exceeds methods that explicitly model ambiguity under disagreement supervision (Probabilistic U-Net, Annotator-Confusion 3D-UNet). A qualitative feature-space analysis shows that presence ambiguity is already encoded in the frozen encoder features of pixel-wise-trained networks, only to be discarded by the segmentation output and its entropy aggregation. Our findings expose a fundamental mismatch between the theoretical promise of aleatoric uncertainty and its practical behavior, and suggest that practitioners should not rely on entropy-based uncertainty as a proxy for clinical ambiguity in safety-critical applications.

Problem

• Aleatoric Uncertainty
• Presence Ambiguity
• 3D Lung Nodule Segmentation
• Entropy-based Uncertainty
• Medical Image Segmentation

Innovation

• Presence Ambiguity
• Aleatoric Uncertainty
• Supervised Ambiguity Head
• Frozen Segmentation Features
• 3D Lung Nodule Segmentation