One paper has been accepted by IEEE TMI 🎉
Title: Semantic Augmentation Variational Autoencoder for Unsupervised Anomaly Detection in Retinal OCT Images
Our paper on unsupervised anomaly detection in retinal OCT images has been accepted by IEEE Transactions on Medical Imaging. This work proposes SeAugVAE, a semantic augmentation variational autoencoder that improves anomaly sensitivity by enforcing dual distribution consistency in both image and feature spaces during VAE training. It further develops structural-semantic anomaly attention maps to localize retinal anomalies from local and global perspectives.
