DOMINO: Domain-aware Model Calibration in Medical Image Segmentation

TL;DR

DOMINO is a novel domain-aware calibration method that improves the reliability and accuracy of medical image segmentation models by leveraging semantic class relationships, especially benefiting rarer classes.

Contribution

Introduces DOMINO, a domain-aware calibration technique utilizing semantic class similarities to enhance model calibration and performance in medical image segmentation.

Findings

Outperforms non-calibrated models and morphometric methods in head image segmentation.

Achieves better calibration, higher accuracy, and faster inference, especially on rarer classes.

Enhances trustworthiness of deep learning models in medical imaging.

Abstract

Model calibration measures the agreement between the predicted probability estimates and the true correctness likelihood. Proper model calibration is vital for high-risk applications. Unfortunately, modern deep neural networks are poorly calibrated, compromising trustworthiness and reliability. Medical image segmentation particularly suffers from this due to the natural uncertainty of tissue boundaries. This is exasperated by their loss functions, which favor overconfidence in the majority classes. We address these challenges with DOMINO, a domain-aware model calibration method that leverages the semantic confusability and hierarchical similarity between class labels. Our experiments demonstrate that our DOMINO-calibrated deep neural networks outperform non-calibrated models and state-of-the-art morphometric methods in head image segmentation. Our results show that our method can consistently achieve better calibration, higher accuracy, and faster inference times than these methods, especially on rarer classes. This performance is attributed to our domain-aware regularization to inform semantic model calibration. These findings show the importance of semantic ties between class labels in building confidence in deep learning models. The framework has the potential to improve the trustworthiness and reliability of generic medical image segmentation models. The code for this article is available at: https://github.com/lab-smile/DOMINO.