Deep Autoencoding Models for Unsupervised Anomaly Segmentation in Brain MR Images

By Christoph Baur, Benedikt Wiestler, Shadi Albarqouni, Nassir Navab. BrainLes@MICCAI, 2018

495 citations recorded by the source corpus.

Reliably modeling normality and differentiating abnormal appearances from normal cases is a very appealing approach for detecting pathologies in medical images. A plethora of such unsupervised anomaly detection approaches has been made in the medical domain, based on statistical methods, content-based retrieval, clustering and recently also deep learning. Previous approaches towards deep unsupervised anomaly detection model patches of normal anatomy with variants of Autoencoders or GANs, and detect anomalies either as outliers in the learned feature space or from large reconstruction errors. In contrast to these patch-based approaches, we show that deep spatial autoencoding models can be efficiently used to capture normal anatomical variability of entire 2D brain MR images. A variety of experiments on real MR data containing MS lesions corroborates our hypothesis that we can detect and even delineate anomalies in brain MR images by simply comparing input images to their reconstruction. Results show that constraints on the latent space and adversarial training can further improve the segmentation performance over standard deep representation learning.

Subfields: Artificial Intelligence, Radiology, Nuclear Medicine and Imaging

View Paper: https://www.semanticscholar.org/paper/eb60fe884c53b420edbce57059b242cfcbae0f7c

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Radiology, Nuclear Medicine and Imaging

Citations in the corpus

Student-Teacher Feature Pyramid Matching for Anomaly Detection (2021, methodology): “The pixel-level methods advance anomaly detection by means of pixel-wise comparison of image patches and their reconstructions [6, 34, 35] or per-pixel estimation of probability density on entire images [1, 37], among which Auto-encoders, Generative Adversarial Networks (GANs), and their variants ar”
MVTec AD — A Comprehensive Real-World Dataset for Unsupervised Anomaly Detection (2019, background): “[3] for the unsupervised segmentation of anomalies in brain MR scans.”
Unsupervised anomaly localization using VAE and beta-VAE (2020, methodology): “Indeed, the inability of the model to re-construct a given part of an image is used as a way to segment the anomaly, using a pixel-wise threshold on the reconstruction error [4,3,5].”
Multiresolution Knowledge Distillation for Anomaly Detection (2020, unknown): “DSVDD[33] Auto-Encoder[55] AnoGan[41] VAE-GAN[5] Pix2Pix[20] GANomaly[3]”
Scaling Out-of-Distribution Detection for Real-World Settings (2022, methodology): “GT is ground truth, the autoencoder model is based on the spatial autoencoder used in (Baur et al., 2019), MSP is the maximum softmax probability baseline (Hendrycks & Gimpel, 2017), and MaxLogit is the method we propose as a new baseline for large-scale settings.”
PyOD: A Python Toolbox for Scalable Outlier Detection (2019, background): “Industry applications include fraud detection in ﬁnance (Ahmed et al., 2016), fault diagnosis in mechanics (Shin et al., 2005), intrusion detection in network security (Garcia-Teodoro et al., 2009) and pathology detection in medical imaging (Baur et al., 2018).”
EfficientAD: Accurate Visual Anomaly Detection at Millisecond-Level Latencies (2023, methodology): “[6] Christoph Baur, Benedikt Wiestler, Shadi Albarqouni, and Nassir Navab.”
Deep learning based brain tumor segmentation: a survey (2020, background): “[74] compared several sets of AEs and VAEs for brain lesion detection.”
Improving Unsupervised Defect Segmentation by Applying Structural Similarity to Autoencoders (2018, methodology): “[16] propose a general framework for defect segmentation using autoencoding architectures and a perpixel error metric based on the `(1)-distance.”
DeSTSeg: Segmentation Guided Denoising Student-Teacher for Anomaly Detection (2022, methodology): “In image reconstruction, researchers used autoencoder [4], variational autoencoder [1, 31] or generative adversarial network [22, 28, 29] to train an image reconstruction model on normal data.”
Uninformed Students: Student-Teacher Anomaly Detection With Discriminative Latent Embeddings (2019, background): “Existing work predominantly focuses on generative algorithms such as Generative Adversarial Networks (GANs) [31, 32] or Variational Autoencoders (VAEs) [5, 36].”
GANs for Medical Image Analysis (2018, background): “Succeeding work in [99] and [100] transferred this concept to anomaly detection in brain MR images.”
Autoencoders for Unsupervised Anomaly Segmentation in Brain MR Images: A Comparative Study (2020, methodology): “[3] leveraged the VAEGAN [16]—a combination of the GAN and VAE (Fig.”
CFLOW-AD: Real-Time Unsupervised Anomaly Detection with Localization via Conditional Normalizing Flows (2021, methodology): “A fully-generative models [35, 36, 3, 38] are directly applied to images in order to estimate pixel-level probability density and compute per-pixel reconstruction errors as”