Our paper “Distribution-Aware Replay for Continual MRI Segmentation” got accepted the Workshop Personalized Incremental Learning in Medicine at MICCAI 2024, with proceedings to be published in Lecture Notes in Computer Science (LNCS), Springer.
In the paper, we demonstrate that a small generative model (a variational autoencoder), that is trained to learn the latent distribution of a medical UNet, can equip the system with both continual learning and out-of-distribution detection capabilities. This in turn helps to make the system more robust and learn over time in the presence of sensor, patient population, or task shifts.
For more information, please read the full paper. The abstract is provided below:
“Medical image distributions shift constantly due to changes in patient population and discrepancies in image acquisition. These distribution changes result in performance deterioration; deterioration that continual learning aims to alleviate. However, only adaptation with data rehearsal strategies yields practically desirable performance for medical image segmentation. Such rehearsal violates patient privacy and, as most continual learning approaches, overlooks unexpected changes from out-of-distribution instances. To transcend both of these challenges, we introduce a distribution-aware replay strategy that mitigates forgetting through auto-encoding of features, while simultaneously leveraging the learned distribution of features to detect model failure. We provide empirical corroboration on hippocampus and prostate MRI segmentation. To ensure reproducibility, we make our code available at https://github.com/MECLabTUDA/Lifelong-nnUNet/tree/