We Care Each Pixel: Calibrating on Medical Segmentation Model

2025-03-07

AI Summary

Medical image segmentation models often suffer from unreliable confidence estimates and a disconnect between spatial accuracy and probabilistic calibration in clinical practice. To address this, we propose the pixel-wise Expected Calibration Error (pECE) as a novel metric for evaluating calibration quality at the pixel level. We introduce the Signed Distance Calibration Loss (SDC), which explicitly incorporates boundary geometry—modeled via signed distance fields—into the calibration objective. Furthermore, we integrate morphological adaptive mask preprocessing with enhanced Margin SVLS/NACL losses to jointly optimize segmentation accuracy and probabilistic calibration. Evaluated across multiple medical segmentation benchmarks, our method achieves significant improvements: +0.8% in Dice score and −32% reduction in pECE, yielding more robust and clinically trustworthy probabilistic predictions.

Technology Category

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

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Security and Privacy: Large-scale security measurements
• User Modeling, Personalization and Recommendation: Metrics for user behavior and evaluating success

Abstract

Medical image segmentation is fundamental for computer-aided diagnostics, providing accurate delineation of anatomical structures and pathological regions. While common metrics such as Accuracy, DSC, IoU, and HD primarily quantify spatial agreement between predictions and ground-truth labels, they do not assess the calibration quality of segmentation models, which is crucial for clinical reliability. To address this limitation, we propose pixel-wise Expected Calibration Error (pECE), a novel metric that explicitly measures miscalibration at the pixel level, thereby ensuring both spatial precision and confidence reliability. We further introduce a morphological adaptation strategy that applies morphological operations to ground-truth masks before computing calibration losses, particularly benefiting margin-based losses such as Margin SVLS and NACL. Additionally, we present the Signed Distance Calibration Loss (SDC), which aligns boundary geometry with calibration objectives by penalizing discrepancies between predicted and ground-truth signed distance functions (SDFs). Extensive experiments demonstrate that our method not only enhances segmentation performance but also improves calibration quality, yielding more trustworthy confidence estimates. Code is available at: https://github.com/EagleAdelaide/SDC-Loss.

Problem

• Assessing calibration quality in medical segmentation models
• Introducing pixel-wise Expected Calibration Error (pECE)
• Improving segmentation performance and calibration reliability

Innovation

• Pixel-wise Expected Calibration Error (pECE)
• Morphological adaptation strategy for calibration
• Signed Distance Calibration Loss (SDC)