StRegA: Unsupervised Anomaly Detection in Brain MRIs using a Compact Context-encoding Variational Autoencoder

  title={StRegA: Unsupervised Anomaly Detection in Brain MRIs using a Compact Context-encoding Variational Autoencoder},
  author={Soumick Chatterjee and Alessandro Sciarra and Max D{\"u}nnwald and Pavan Kumar Tummala and Shubham Agrawal and Aishwarya Jauhari and Aman Kalra and Steffen Oeltze-Jafra and Oliver Speck and A. N{\"u}rnberger},
  journal={Computers in biology and medicine},



Context-encoding Variational Autoencoder for Unsupervised Anomaly Detection

This work proposes the Context-encoding Variational Autoencoder (ceVAE) which combines reconstruction- with density-based anomaly scoring and achieves unsupervised ROC-AUCs of 0.95 and 0.89, thus outperforming state-of-the-art methods by a considerable margin.

Anomaly Detection Through Latent Space Restoration Using Vector Quantized Variational Autoencoders

An out-of-distribution detection method that combines density and restoration-based approaches using Vector-Quantized Variational Auto-Encoders (VQ-VAEs) finds that the prior probability estimated by the AR model can be useful for unsupervised anomaly detection and enables the estimation of both sample and pixel-wise anomaly scores.

Neural Discrete Representation Learning

Pairing these representations with an autoregressive prior, the model can generate high quality images, videos, and speech as well as doing high quality speaker conversion and unsupervised learning of phonemes, providing further evidence of the utility of the learnt representations.

Unsupervised brain imaging 3D anomaly detection and segmentation with transformers

Anomaly Detection With Representative Neighbors.

An effective anomaly detection method with representativeNeighborhood information is transformed into similarity relations, making the data converge or disperse and anomalies are discriminated by a tailored graph clustering technique, which can effectively reveal structural information of the data.