Rethinking Autoencoders for Medical Anomaly Detection from A Theoretical Perspective

arXiv:2403.09303v1 Announce Type: new
Abstract: Medical anomaly detection aims to identify abnormal findings using only normal training data, playing a crucial role in health screening and recognizing rare diseases. Reconstruction-based methods, particularly those utilizing autoencoders (AEs), are dominant in this field. They work under the assumption that AEs trained on only normal data cannot reconstruct unseen abnormal regions well, thereby enabling the anomaly detection based on reconstruction errors. However, this assumption does not always hold due to the mismatch between …

abstract anomaly anomaly detection arxiv autoencoders cs.cv cs.lg data detection diseases health identify medical normal perspective playing rare diseases role screening training training data type work