Identifying and Categorizing Anomalies in Retinal Imaging Data
This work addresses the challenge of scaling anomaly detection in medical imaging for clinical diagnosis, though it is incremental as it builds on existing autoencoder and one-class SVM methods.
The paper tackled the problem of identifying and categorizing anomalies in retinal imaging data without relying on expert annotations, using healthy examples as a reference, and demonstrated the ability to detect pathologic regions and categorize findings into clinically meaningful classes.
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.