MIM-OOD: Generative Masked Image Modelling for Out-of-Distribution Detection in Medical Images
Preprint
Abstract
Unsupervised Out-of-Distribution (OOD) detection consists in identifying anomalous regions in images leveraging only models trained on images of healthy anatomy. An established approach is to tokenize images and model the distribution of tokens with Auto-Regressive (AR) models. AR models are used to 1) identify anomalous tokens and 2) in-paint anomalous representations with in-distribution tokens. However, AR models are slow at inference time and prone to error accumulation issues which negatively affect OOD detection performance. Our novel method, MIM-OOD, overcomes both speed and error accumulation issues by replacing the AR model with two task-specific networks: 1) a transformer optimized to identify anomalous tokens and 2) a transformer optimized to in-paint anomalous tokens using masked image modelling (MIM). Our experiments with brain MRI anomalies show that MIM-OOD substantially outperforms AR models (DICE 0.458 vs 0.301) while achieving a nearly 25x speedup (9.5s vs 244s).Keywords
All Related Versions
This publication has 9 references indexed in Scilit:
- Fast Unsupervised Brain Anomaly Detection and Segmentation with Diffusion Models2022
- Unsupervised brain imaging 3D anomaly detection and segmentation with transformersMedical Image Analysis, 2022
- Detecting Outliers with Foreign Patch InterpolationMachine Learning for Biomedical Imaging, 2022
- Autoencoders for unsupervised anomaly segmentation in brain MR images: A comparative studyMedical Image Analysis, 2021
- Image Anomaly Detection Using Normal Data Only by Latent Space ResamplingApplied Sciences, 2020
- f-AnoGAN: Fast unsupervised anomaly detection with generative adversarial networksMedical Image Analysis, 2019
- A survey on deep learning in medical image analysisMedical Image Analysis, 2017
- The Multimodal Brain Tumor Image Segmentation Benchmark (BRATS)IEEE Transactions on Medical Imaging, 2015
- The Human Connectome Project: A data acquisition perspectiveNeuroImage, 2012