Learning from pseudo-lesion: a self-supervised framework for COVID-19 diagnosis

The Coronavirus disease 2019 (COVID-19) has rapidly spread all over the world since its first report in December 2019, and thoracic computed tomography (CT) has become one of the main tools for its diagnosis. In recent years, deep learning-based approaches have shown impressive performance in myriad image recognition tasks. However, they usually require a large number of annotated data for training. Inspired by ground glass opacity, a common finding in COIVD-19 patient's CT scans, we proposed in this paper a novel self-supervised pretraining method based on pseudo-lesion generation and restoration for COVID-19 diagnosis. We used Perlin noise, a gradient noise based mathematical model, to generate lesion-like patterns, which were then randomly pasted to the lung regions of normal CT images to generate pseudo-COVID-19 images. The pairs of normal and pseudo-COVID-19 images were then used to train an encoder-decoder architecture-based U-Net for image restoration, which does not require any labeled data. The pretrained encoder was then fine-tuned using labeled data for COVID-19 diagnosis task. Two public COVID-19 diagnosis datasets made up of CT images were employed for evaluation. Comprehensive experimental results demonstrated that the proposed self-supervised learning approach could extract better feature representation for COVID-19 diagnosis, and the accuracy of the proposed method outperformed the supervised model pretrained on large-scale images by 6.57% and 3.03% on SARS-CoV-2 dataset and Jinan COVID-19 dataset, respectively.

Automated summary

In this paper, a novel self-supervised pretraining method based on pseudo-lesion generation and restoration was proposed for COVID-19 diagnosis. The method trained an encoder-decoder architecture-based U-Net using pairs of pseudo-COVID-19 images and normal CT images for image restoration, and then fine-tuned the pretrained encoder using labeled data. Experimental results demonstrated that the proposed method extracted better feature representation for COVID-19 diagnosis, achieving higher accuracy compared to the supervised model pretrained on large-scale images by 6.57% and 3.03% on SARS-CoV-2 dataset and Jinan COVID-19 dataset, respectively.