Quasi-Binarized Autoencoders: An Architecture-Independent Information Bottleneck for Medical Image Anomaly Detection

Key points

• Reconstruction-based methods pass an image through an encoder-decoder network trained on normal data and detect anomalies from the residual between the image and its reconstruction.1
• This works only if the information passed from the encoder to the decoder is limited; otherwise the network learns an identity mapping and reconstructs anomalies too.1
• We introduce the quasi-binarizing (QB) layer, which squashes each latent element into [0, 1] and adds Laplace noise of scale 1/epsilon.1
• Placing a QB layer on every encoder-decoder path, including all skip connections, we build QBAE, a seven-level attention U-Net with 32,768 QB elements.1

Sources (1)

• [1]Quasi-Binarized Autoencoders: An Architecture-Independent Information Bottleneck for Medical Image Anomaly Detection
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 08:34 AM

  Unsupervised anomaly detection, which learns only from normal images, is a central task in medical image analysis and remains an open problem.

  Reconstruction-based methods pass an image through an encoder-decoder network trained on normal data and detect anomalies from the residual between the image and its reconstruction.

Extractive summary: sentences quoted from the sources.