Conceptual
Browse Instruct Curate

SE-ADA: Domain-Invariant Feature Learning for Content-Based Brain MRI Retrieval

A domain-adaptation method (style encoder adversarial domain adaptation) that learns a low-dimensional representation of 3D brain MR images invariant to the imaging site while preserving pathological features, enabling content-based image retrieval across scanners and protocols. A dedicated style encoder isolates domain-specific information and adversarial learning removes it, scaling to unknown domains without a per-domain decoder.