Deep image mining for diabetic retinopathy screening

Abstract

HighlightsA ConvNet is trained to detect images with referable diabetic retinopathy (DR).Training is supervised with image‐level annotations only (referable DR or not).Thanks to a modified training procedure, pathological pixels can also be detected.This solution outperforms DR screening systems in a public dataset of 90,000 images.It also outperforms lesion detectors in the DiaretDB1 dataset, for all lesion types. Graphical abstract Figure. No Caption available. ABSTRACT Deep learning is quickly becoming the leading methodology for medical image analysis. Given a large medical archive, where each image is associated with a diagnosis, efficient pathology detectors or classifiers can be trained with virtually no expert knowledge about the target pathologies. However, deep learning algorithms, including the popular ConvNets, are black boxes: little is known about the local patterns analyzed by ConvNets to make a decision at the image level. A solution is proposed in this paper to create heatmaps showing which pixels in images play a role in the image‐level predictions. In other words, a ConvNet trained for image‐level classification can be used to detect lesions as well. A generalization of the backpropagation method is proposed in order to train ConvNets that produce high‐quality heatmaps. The proposed solution is applied to diabetic retinopathy (DR) screening in a dataset of almost 90,000 fundus photographs from the 2015 Kaggle Diabetic Retinopathy competition and a private dataset of almost 110,000 photographs (e‐ophtha). For the task of detecting referable DR, very good detection performance was achieved: Symbol in Kaggle’s dataset and Symbol in e‐ophtha. Performance was also evaluated at the image level and at the lesion level in the DiaretDB1 dataset, where four types of lesions are manually segmented: microaneurysms, hemorrhages, exudates and cotton‐wool spots. For the task of detecting images containing these four lesion types, the proposed detector, which was trained to detect referable DR, outperforms recent algorithms trained to detect those lesions specifically, with pixel‐level supervision. At the lesion level, the proposed detector outperforms heatmap generation algorithms for ConvNets. This detector is part of the Messidor® system for mobile eye pathology screening. Because it does not rely on expert knowledge or manual segmentation for detecting relevant patterns, the proposed solution is a promising image mining tool, which has the potential to discover new biomarkers in images. Symbol. No Caption available. Symbol. No Caption available.