Towards Reliable Medical Image Segmentation by utilizing Evidential Calibrated Uncertainty

Ke Zou,Yidi Chen,Ling Huang,Xuedong Yuan,Xiaojing Shen,Meng Wang,Rick Siow Mong Goh,Yong Liu,Huazhu Fu

2024-04-14

Abstract:Medical image segmentation is critical for disease diagnosis and treatment assessment. However, concerns regarding the reliability of segmentation regions persist among clinicians, mainly attributed to the absence of confidence assessment, robustness, and calibration to accuracy. To address this, we introduce DEviS, an easily implementable foundational model that seamlessly integrates into various medical image segmentation networks. DEviS not only enhances the calibration and robustness of baseline segmentation accuracy but also provides high-efficiency uncertainty estimation for reliable predictions. By leveraging subjective logic theory, we explicitly model probability and uncertainty for the problem of medical image segmentation. Here, the Dirichlet distribution parameterizes the distribution of probabilities for different classes of the segmentation results. To generate calibrated predictions and uncertainty, we develop a trainable calibrated uncertainty penalty. Furthermore, DEviS incorporates an uncertainty-aware filtering module, which utilizes the metric of uncertainty-calibrated error to filter reliable data within the dataset. We conducted validation studies to assess both the accuracy and robustness of DEviS segmentation, along with evaluating the efficiency and reliability of uncertainty estimation. These evaluations were performed using publicly available datasets including ISIC2018, LiTS2017, and BraTS2019. Additionally, two potential clinical trials are being conducted at Johns Hopkins OCT, Duke-OCT-DME, and FIVES datasets to demonstrate their efficacy in filtering high-quality or out-of-distribution data. Our code has been released in <a class="link-external link-https" href="https://github.com/Cocofeat/DEviS" rel="external noopener nofollow">this https URL</a>.

Image and Video Processing,Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The paper aims to address the reliability issues in the field of medical image segmentation, with a particular focus on how to improve the accuracy, robustness, and credibility of model predictions. Specifically, the researchers developed a method called DEviS (Deep Evidential Segmentation), which can be seamlessly integrated into various medical image segmentation networks, and not only improves the calibration and robustness of baseline segmentation accuracy but also provides efficient uncertainty estimation to achieve reliable predictions.

The key contributions of the paper include:
1. **Probabilistic and Uncertainty Modeling**: Through the theory of Subjective Logic, probabilities and uncertainties for different categories of segmentation results are explicitly modeled. The probability distribution of different category segmentation results is parameterized using the Dirichlet distribution.
2. **Trainable Calibration Uncertainty Penalty (CUP)**: A trainable Calibration Uncertainty Penalty mechanism is developed to generate more calibrated confidence levels while maintaining the segmentation performance of the baseline network.
3. **Uncertainty-Aware Filtering (UAF) Strategy**: An Uncertainty-Aware Filtering strategy is proposed, which uses uncertainty calibration errors to filter reliable data, aiding in the application of DEviS to clinical scenarios, such as detecting anomalous data and indicating high-quality data.

The paper validates the effectiveness of DEviS through experiments on public datasets such as ISIC2018, LiTS2017, and BraTS2019, assessing its accuracy and robustness in predicting standard and adversarial samples (such as noise, blur, or random masks), as well as the efficiency and reliability of uncertainty estimation. Furthermore, the researchers also explored the potential of DEviS in clinical applications, including applications in anomaly data detection and image quality metrics.

In summary, the paper introduces the DEviS method with the aim of enhancing the reliability and practicality of medical image segmentation tasks, especially in clinical practice applications.