Identifying and Categorizing Anomalies in Retinal Imaging Data

Abstract

The identification and quantification of markers in medical images is critical for diagnosis, prognosis and management of patients in clinical practice. Supervised- or weakly supervised training enables the detection of findings that are known a priori. It does not scale well, and a priori definition limits the vocabulary of markers to known entities reducing the accuracy of diagnosis and prognosis. Here, we propose the identification of anomalies in large-scale medical imaging data using healthy examples as a reference. We detect and categorize candidates for anomaly findings untypical for the observed data. A deep convolutional autoencoder is trained on healthy retinal images. The learned model generates a new feature representation, and the distribution of healthy retinal patches is estimated by a one-class support vector machine. Results demonstrate that we can identify pathologic regions in images without using expert annotations. A subsequent clustering categorizes findings into clinically meaningful classes. In addition the learned features outperform standard embedding approaches in a classification task.

Extended Abstract, Accepted for NIPS 2016 Workshop "Machine Learning for Health"

References in corpus (4)

• Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs)
• Unsupervised Visual Representation Learning by Context Prediction
• Stacked What-Where Auto-encoders
• Unsupervised Object Discovery and Localization in the Wild: Part-based Matching with Bottom-up Region Proposals

Cited by in corpus (13)

• Deep Autoencoding Models for Unsupervised Anomaly Segmentation in Brain MR Images
• Unsupervised Anomaly Detection with Generative Adversarial Networks to Guide Marker Discovery
• Unsupervised phase discovery with deep anomaly detection
• Student-Teacher Feature Pyramid Matching for Anomaly Detection
• Unsupervised Anomaly Detection with an Enhanced Teacher for Student-Teacher Feature Pyramid Matching
• Rethinking Assumptions in Deep Anomaly Detection
• Robust Image Segmentation Quality Assessment
• DROCC: Deep Robust One-Class Classification
• Focus Your Distribution: Coarse-to-Fine Non-Contrastive Learning for Anomaly Detection and Localization
• Meta Learning for Few-Shot One-class Classification
• An Alarm System For Segmentation Algorithm Based On Shape Model
• Progressive GANomaly: Anomaly detection with progressively growing GANs
• DDR-ID: Dual Deep Reconstruction Networks Based Image Decomposition for Anomaly Detection