Domain Adaptation Using Pseudo Labels for COVID-19 Detection

TL;DR

This work tackles domain shift in COVID-19 detection from CT scans under limited annotations by introducing a two-stage domain adaptation framework that uses pseudo labels to leverage labeled data from one domain and unlabeled data from another. Stage 1 trains on annotated data from both domains with augmentation and contrastive learning, while Stage 2 generates pseudo labels for unlabeled Domain B data to retrain the model, enhancing cross-domain performance via a CMC-based backbone. The framework achieves a Macro F1 score of 0.92 on the Covid-19 domain adaptation validation set, surpassing a Monte Carlo Dropout baseline, and demonstrates improved generalization across institutions. These results suggest a scalable, efficient approach for rapid COVID-19 screening in diverse clinical settings, potentially reducing annotation needs and healthcare burdens.

Abstract

In response to the need for rapid and accurate COVID-19 diagnosis during the global pandemic, we present a two-stage framework that leverages pseudo labels for domain adaptation to enhance the detection of COVID-19 from CT scans. By utilizing annotated data from one domain and non-annotated data from another, the model overcomes the challenge of data scarcity and variability, common in emergent health crises. The innovative approach of generating pseudo labels enables the model to iteratively refine its learning process, thereby improving its accuracy and adaptability across different hospitals and medical centres. Experimental results on COV19-CT-DB database showcase the model's potential to achieve high diagnostic precision, significantly contributing to efficient patient management and alleviating the strain on healthcare systems. Our method achieves 0.92 Macro F1 Score on the validation set of Covid-19 domain adaptation challenge.

Figures (1)

• Figure 1: Overview of our framework for COVID-19 domain adaptation.