Diabetic Retinopathy Fundus Image Classification and Lesions Localization System Using Deep Learning

Abstract

Diabetic retinopathy (DR) is a disease resulting from diabetes complications, causing non-reversible damage to retina blood vessels. DR is a leading cause of blindness if not detected early. The currently available DR treatments are limited to stopping or delaying the deterioration of sight, highlighting the importance of regular scanning using high-efficiency computer-based systems to diagnose cases early. The current work presented fully automatic diagnosis systems that exceed manual techniques to avoid misdiagnosis, reducing time, effort and cost. The proposed system classifies DR images into five stages-no-DR, mild, moderate, severe and proliferative DR-as well as localizing the affected lesions on retain surface. The system comprises two deep learning-based models. The first model (CNN512) used the whole image as an input to the CNN model to classify it into one of the five DR stages. It achieved an accuracy of 88.6% and 84.1% on the DDR and the APTOS Kaggle 2019 public datasets, respectively, compared to the state-of-the-art results. Simultaneously, the second model used an adopted YOLOv3 model to detect and localize the DR lesions, achieving a 0.216 mAP in lesion localization on the DDR dataset, which improves the current state-of-the-art results. Finally, both of the proposed structures, CNN512 and YOLOv3, were fused to classify DR images and localize DR lesions, obtaining an accuracy of 89% with 89% sensitivity, 97.3 specificity and that exceeds the current state-of-the-art results.

Keywords: YOLO; computer-aided diagnosis; convolutional neural networks; deep learning; diabetic retinopathy; diabetic retinopathy classification; diabetic retinopathy lesions localization.

Figure 1

The different types of DR lesions.

Figure 2

The DR stages: (a) No DR (b) Mild, (c) Moderate, (d) Severe, (e) Proliferative DR.

Figure 3

The ratio of studies that classified the DR stages [47].

Figure 4

The ratio of studies that classified the DR lesions [47].

Figure 5

Block diagram of the different proposed models for DR images classification and localization.

Figure 6

The retina images preprocessing methods.

Figure 7

Sample images of the (a) original image and (b) the preprocessing image.

Figure 8

Sample of an image augmentation: (a) original image, (b) flipped image, (c) rotated image, (d) sheared image, (e) translated image and (f) brightened image.

Figure 9

The proposed custom CNN architectures: (a) CNN299 and (b) CNN512.

Figure 10

Transfer learning EfficientNetB0.

Figure 11

The proposed Lesion Localization method to detect DR stages and locate lesions for (a) train and (b) test images.

Figure 12

The ROC curves of the (a) APTOS 2019 and (b) the DDR datasets on CNN512.

Figure 13

Sample of the DDR images visualization for: (a) the ground truth images annotation, (b) predicted images by fused model.

Figure 14

The confusion matrixes and ROC curves of the DDR dataset on fused models.

Figure 15

Samples of the miss labeled lesions from the DDR dataset compared to the predicted lesions by YOLOv3.