Multi-label classification of retinal lesions in diabetic retinopathy for automatic analysis of fundus fluorescein angio
- PDF / 1,528,826 Bytes
- 7 Pages / 595.276 x 790.866 pts Page_size
- 68 Downloads / 203 Views
RETINAL DISORDERS
Multi-label classification of retinal lesions in diabetic retinopathy for automatic analysis of fundus fluorescein angiography based on deep learning Xiangji Pan 1 & Kai Jin 1 & Jing Cao 1 & Zhifang Liu 1 & Jian Wu 2 & Kun You 2 & Yifei Lu 2 & Yufeng Xu 1 & Zhaoan Su 1 & Jiekai Jiang 1 & Ke Yao 1 & Juan Ye 1 Received: 12 September 2019 / Revised: 30 October 2019 / Accepted: 13 December 2019 # Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract Purpose To automatically detect and classify the lesions of diabetic retinopathy (DR) in fundus fluorescein angiography (FFA) images using deep learning algorithm through comparing 3 convolutional neural networks (CNNs). Methods A total of 4067 FFA images from Eye Center at the Second Affiliated Hospital of Zhejiang University School of Medicine were annotated with 4 kinds of lesions of DR, including non-perfusion regions (NP), microaneurysms, leakages, and laser scars. Three CNNs including DenseNet, ResNet50, and VGG16 were trained to achieve multi-label classification, which means the algorithms could identify 4 retinal lesions above at the same time. Results The area under the curve (AUC) of DenseNet reached 0.8703, 0.9435, 0.9647, and 0.9653 for detecting NP, microaneurysms, leakages, and laser scars, respectively. For ResNet50, AUC was 0.8140 for NP, 0.9097 for microaneurysms, 0.9585 for leakages, and 0.9115 for laser scars. And for VGG16, AUC was 0.7125 for NP, 0.5569 for microaneurysms, 0.9177 for leakages, and 0.8537 for laser scars. Conclusions Experimental results demonstrate that DenseNet is a suitable model to automatically detect and distinguish retinal lesions in the FFA images with multi-label classification, which lies the foundation of automatic analysis for FFA images and comprehensive diagnosis and treatment decision-making for DR. Keywords Diabetic retinopathy . Fundus fluorescein angiography . Deep learning . Multi-label classification
Introduction Diabetic retinopathy (DR) is a leading cause of irreversible vision loss in working-age adults(20 to 65 years) and approximately one in three people living with diabetes have some degree of DR and one in ten will develop a vision-threatening form of the disease. As the estimates of the International Association on the Prevention of Blindness (IAPB), 145 mil-
Xiangji Pan, Kai Jin and Jing Cao contributed equally to this work. * Juan Ye [email protected] 1
Department of Ophthalmology, The Second Affiliated Hospital of Zhejiang University, College of Medicine, Hangzhou 310009, China
2
College of Computer Science and Technology, Zhejiang University, Hangzhou 310027, China
lion people had some form of DR, and 45 million people suffered from vision-threatening DR in 2015 [1]. According to an abrupt rise in diabetes, the increasing burden of DR can be forecasted [2]. Early detection and intervention of DR is of great importance. The common diagnostic methods for DR include retinal fundus photography, optical coherence tomography (OCT), and fundus fluorescein angiography (FFA).
Data Loading...
