# Privacy-Aware Continual Self-Supervised Learning on Multi-Window Chest Computed Tomography for Domain-Shift Robustness

## TL;DR

This paper introduces a new self-supervised learning framework for chest CT images that improves model robustness to domain shifts while preserving data privacy.

## Contribution

The novel contribution is a privacy-aware continual self-supervised learning framework with latent replay and feature distillation for domain-shift robustness in chest CT.

## Key findings

- The proposed method outperforms existing approaches in handling domain shifts in chest CT images.
- Latent replay and feature distillation techniques effectively mitigate catastrophic forgetting and improve model generalizability.
- The framework maintains data privacy while learning from unlabeled images across different window settings.

## Abstract

We propose a novel continual self-supervised learning (CSSL) framework for simultaneously learning diverse features from multi-window-obtained chest computed tomography (CT) images and ensuring data privacy. Achieving a robust and highly generalizable model in medical image diagnosis is challenging, mainly because of issues, such as the scarcity of large-scale, accurately annotated datasets and domain shifts inherent to dynamic healthcare environments. Specifically, in chest CT, these domain shifts often arise from differences in window settings, which are optimized for distinct clinical purposes. Previous CSSL frameworks often mitigated domain shift by reusing past data, a typically impractical approach owing to privacy constraints. Our approach addresses these challenges by effectively capturing the relationship between previously learned knowledge and new information across different training stages through continual pretraining on unlabeled images. Specifically, by incorporating a latent replay-based mechanism into CSSL, our method mitigates catastrophic forgetting due to domain shifts during continual pretraining while ensuring data privacy. Additionally, we introduce a feature distillation technique that integrates Wasserstein distance-based knowledge distillation and batch-knowledge ensemble, enhancing the ability of the model to learn meaningful, domain-shift-robust representations. Finally, we validate our approach using chest CT images obtained across two different window settings, demonstrating superior performance compared with other approaches.