Sparse-GAN: Sparsity-constrained Generative Adversarial Network for Anomaly Detection in Retinal OCT Image

TL;DR

Sparse-GAN introduces a sparsity-constrained generative adversarial network that detects retinal anomalies in OCT images using only healthy data, improving disease screening by predicting anomalies in latent space and highlighting lesion regions.

Contribution

The paper presents a novel anomaly detection framework that predicts anomalies in latent space with sparsity regularization, enabling disease detection without diseased training data.

Findings

Outperforms state-of-the-art methods on public dataset

Effective in detecting unseen retinal diseases

Provides lesion heatmaps for interpretability

Abstract

With the development of convolutional neural network, deep learning has shown its success for retinal disease detection from optical coherence tomography (OCT) images. However, deep learning often relies on large scale labelled data for training, which is oftentimes challenging especially for disease with low occurrence. Moreover, a deep learning system trained from data-set with one or a few diseases is unable to detect other unseen diseases, which limits the practical usage of the system in disease screening. To address the limitation, we propose a novel anomaly detection framework termed Sparsity-constrained Generative Adversarial Network (Sparse-GAN) for disease screening where only healthy data are available in the training set. The contributions of Sparse-GAN are two-folds: 1) The proposed Sparse-GAN predicts the anomalies in latent space rather than image-level; 2) Sparse-GAN is constrained by a novel Sparsity Regularization Net. Furthermore, in light of the role of lesions for disease screening, we present to leverage on an anomaly activation map to show the heatmap of lesions. We evaluate our proposed Sparse-GAN on a publicly available dataset, and the results show that the proposed method outperforms the state-of-the-art methods.