Uncertainty-Based Ensemble Learning in CMR Semantic Segmentation

TL;DR

This paper presents an uncertainty-based ensemble learning framework called Streaming for cardiac MRI segmentation, improving end-slice accuracy and overall performance, and introduces the End Coefficient for quantifying end-slice accuracy.

Contribution

The authors propose a novel ensemble method that leverages segmentation uncertainty to enhance cardiac MRI segmentation, especially at the challenging end slices, and provide an open-source implementation.

Findings

Achieves near state-of-the-art overall segmentation accuracy.

Outperforms existing models on end-slice segmentation performance.

Provides a new metric, End Coefficient, for quantifying end-slice accuracy.

Abstract

Existing methods derive clinical functional metrics from ventricular semantic segmentation in cardiac cine sequences. While performing well on overall segmentation, they struggle with the end slices. To address this, we extract global uncertainty from segmentation variance and use it in our ensemble learning method, Streaming, for classifier weighting, balancing overall and end-slice performance. We introduce the End Coefficient (EC) to quantify end-slice accuracy. Experiments on ACDC and M\&Ms datasets show that our framework achieves near state-of-the-art Dice Similarity Coefficient (DSC) and outperforms all models on end-slice performance, improving patient-specific segmentation accuracy. We open-sourced our code on https://github.com/LEw1sin/Uncertainty-Ensemble.