Automated Chest CT Image Segmentation of COVID-19 Lung Infection based on 3D U-Net

Dominik Müller,Iñaki Soto Rey,Frank Kramer

DOI: https://doi.org/10.1016/j.imu.2021.100681

2020-06-25

Abstract:The coronavirus disease 2019 (COVID-19) affects billions of lives around the world and has a significant impact on public healthcare. Due to rising skepticism towards the sensitivity of RT-PCR as screening method, medical imaging like computed tomography offers great potential as alternative. For this reason, automated image segmentation is highly desired as clinical decision support for quantitative assessment and disease monitoring. However, publicly available COVID-19 imaging data is limited which leads to overfitting of traditional approaches. To address this problem, we propose an innovative automated segmentation pipeline for COVID-19 infected regions, which is able to handle small datasets by utilization as variant databases. Our method focuses on on-the-fly generation of unique and random image patches for training by performing several preprocessing methods and exploiting extensive data augmentation. For further reduction of the overfitting risk, we implemented a standard 3D U-Net architecture instead of new or computational complex neural network architectures. Through a 5-fold cross-validation on 20 CT scans of COVID-19 patients, we were able to develop a highly accurate as well as robust segmentation model for lungs and COVID-19 infected regions without overfitting on the limited data. Our method achieved Dice similarity coefficients of 0.956 for lungs and 0.761 for infection. We demonstrated that the proposed method outperforms related approaches, advances the state-of-the-art for COVID-19 segmentation and improves medical image analysis with limited data. The code and model are available under the following link: <a class="link-external link-https" href="https://github.com/frankkramer-lab/covid19.MIScnn" rel="external noopener nofollow">this https URL</a>

Image and Video Processing,Computer Vision and Pattern Recognition,Machine Learning

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to achieve high - precision automatic segmentation on a limited COVID - 19 chest CT image data set to support clinical decision - making, especially in terms of disease quantitative assessment and monitoring. Specifically, the authors propose an innovative automated segmentation pipeline, which aims to handle small data sets and reduce the risk of over - fitting by using a variant database to generate unique random image patches for training. In addition, they adopt the standard 3D U - Net architecture instead of new or computationally complex neural network architectures, further reducing the risk of over - fitting. Through 5 - fold cross - validation on CT scans of 20 COVID - 19 patients, they develop a highly accurate and robust model for segmenting the lungs and COVID - 19 - infected areas without over - fitting on the limited data.

The main contributions of the paper include:
- **Data augmentation**: Utilize a wide range of online data augmentation techniques to generate unique random image patches to increase the diversity of training data.
- **3D U - Net architecture**: Select the standard 3D U - Net architecture to avoid the problem of excessive parameters brought by complex architectures.
- **Performance evaluation**: Through 5 - fold cross - validation, demonstrate the robustness and accuracy of the model on a limited data set.

These methods together solve the challenge of training a high - precision segmentation model on a limited data set, especially in medical image analysis where data acquisition is difficult and annotation is time - consuming and labor - intensive.