Towards Synchronous Memorizability and Generalizability with Site-Modulated Diffusion Replay for Cross-Site Continual...

The ability to learn sequentially from different data sites is crucial for a deep network in solving practical medical image diagnosis problems due to privacy restrictions and storage limitations. However, adapting to the incoming site leads to catastrophic forgetting on past sites and decreases generalizability on unseen sites. Existing Continual Learning (CL) and Domain Generalization (DG) methods have been proposed to solve these two challenges respectively, but none of them can address both simultaneously. Recognizing this limitation, this paper proposes a novel training paradigm, learning towards Synchronous Memorizability and Generalizability (SMG-Learning). To achieve this, we create the orientational gradient alignment to ensure memorizability on previous sites, and arbitrary gradient alignment to enhance generalizability on unseen sites. This approach is called Parallel Gradient Alignment (PGA). Furthermore, we approximate the PGA as dual meta-objectives using the first-order Taylor expansion to reduce the computational cost of aligning gradients. Considering that performing gradient alignments, especially for previous sites, is not feasible due to privacy constraints, we design a Site-Modulated Diffusion (SMD) model to generate images with site-specific learnable prompts, replaying images have similar data distributions as previous sites. We evaluate our method on two medical image segmentation tasks, where data from different sites arrive sequentially. Experimental results show that our method efficiently enhances both memorizability and generalizability better than other state-of-the-art methods, delivering satisfactory performance across all sites. Our code will be available at: https://github.com/dyxu-cuhkcse/SMG-Learning.