A Deep Ensemble Learning Approach to Lung CT Segmentation for COVID-19 Severity Assessment

TL;DR

This paper introduces a hierarchical deep ensemble learning method for detailed lung CT segmentation in COVID-19 patients, achieving high accuracy, uncertainty estimation, and promising clinical correlation, ranked second in a Kaggle challenge.

Contribution

The novel end-to-end hierarchical network with ensemble learning enhances segmentation accuracy and uncertainty estimation for COVID-19 lung pathologies.

Findings

Achieved competitive segmentation results on three datasets.

Provided uncertainty measures correlating with radiologist disagreements.

Showed preliminary links between segmentation and COVID-19 severity scores.

Abstract

We present a novel deep learning approach to categorical segmentation of lung CTs of COVID-19 patients. Specifically, we partition the scans into healthy lung tissues, non-lung regions, and two different, yet visually similar, pathological lung tissues, namely, ground-glass opacity and consolidation. This is accomplished via a unique, end-to-end hierarchical network architecture and ensemble learning, which contribute to the segmentation and provide a measure for segmentation uncertainty. The proposed framework achieves competitive results and outstanding generalization capabilities for three COVID-19 datasets. Our method is ranked second in a public Kaggle competition for COVID-19 CT images segmentation. Moreover, segmentation uncertainty regions are shown to correspond to the disagreements between the manual annotations of two different radiologists. Finally, preliminary promising correspondence results are shown for our private dataset when comparing the patients' COVID-19 severity scores (based on clinical measures), and the segmented lung pathologies. Code and data are available at our repository: https://github.com/talbenha/covid-seg