Multi-Source COVID-19 Detection via Variance Risk Extrapolation

TL;DR

This paper introduces a domain-invariant training approach for COVID-19 detection from chest CT scans across multiple hospitals, using Variance Risk Extrapolation and Mixup augmentation to improve cross-domain generalization.

Contribution

The paper proposes integrating Variance Risk Extrapolation with Mixup data augmentation to enhance model robustness and generalization across diverse medical imaging sources.

Findings

Achieved an average macro F1 score of 0.96 on validation data.

Reduced overfitting to center-specific features.

Enhanced model robustness through domain-invariant learning.

Abstract

We present our solution for the Multi-Source COVID-19 Detection Challenge, which aims to classify chest CT scans into COVID and Non-COVID categories across data collected from four distinct hospitals and medical centers. A major challenge in this task lies in the domain shift caused by variations in imaging protocols, scanners, and patient populations across institutions. To enhance the cross-domain generalization of our model, we incorporate Variance Risk Extrapolation (VREx) into the training process. VREx encourages the model to maintain consistent performance across multiple source domains by explicitly minimizing the variance of empirical risks across environments. This regularization strategy reduces overfitting to center-specific features and promotes learning of domain-invariant representations. We further apply Mixup data augmentation to improve generalization and robustness. Mixup interpolates both the inputs and labels of randomly selected pairs of training samples, encouraging the model to behave linearly between examples and enhancing its resilience to noise and limited data. Our method achieves an average macro F1 score of 0.96 across the four sources on the validation set, demonstrating strong generalization.