Adversarial Continual Learning for Multi-domain Hippocampal Segmentation

Deep learning for medical imaging suffers from temporal and privacy-related restrictions on data availability. To still obtain viable models, continual aims to train in sequential order, as when is available. The main challenge that methods face prevent catastrophic forgetting, i.e., a decrease performance the encountered earlier. This issue makes continuous training of segmentation models applications extremely difficult. Yet, often, at least two different domains available which we can exploit model way it disregards domain-specific information. We propose an architecture leverages simultaneous availability or more datasets learn disentanglement between content domain adversarial fashion. domain-invariant representation then lays base semantic segmentation. Our approach takes inspiration adaptation combines with hippocampal brain magnetic resonance (MRI). showcase our method reduces forgetting outperforms state-of-the-art methods.