Automatic glaucoma classification using color fundus images based on convolutional neural networks and transfer learning

Abstract

Glaucoma detection in color fundus images is a challenging task that requires expertise and years of practice. In this study we exploited the application of different Convolutional Neural Networks (CNN) schemes to show the influence in the performance of relevant factors like the data set size, the architecture and the use of transfer learning vs newly defined architectures. We also compared the performance of the CNN based system with respect to human evaluators and explored the influence of the integration of images and data collected from the clinical history of the patients. We accomplished the best performance using a transfer learning scheme with VGG19 achieving an AUC of 0.94 with sensitivity and specificity ratios similar to the expert evaluators of the study. The experimental results using three different data sets with 2313 images indicate that this solution can be a valuable option for the design of a computer aid system for the detection of glaucoma in large-scale screening programs.

REFUGE Challenge: A unified framework for evaluating automated methods for glaucoma assessment from fundus photographs

“…Christopher et al (2018) fine-tuned a network initialized with weights learned from ImageNet to detect glaucomatous optic neuropathy. Similarly, transfer learning was shown by Gómez-Valverde et al (2019) to outperform networks trained from scratch for glaucoma detection. Both studies applied a massive image data set with more than 14.000 images to fine tune these networks.…”

“…Most of the literature in glaucoma classification uses receiver-operating characteristic (ROC) curves (Davis and Goadrich, 2006) for evaluation, including the area under the curve (AUC) as a summary value (Chen et al, 2015a,b;Christopher et al, 2018;Fu et al, 2018;Gómez-Valverde et al, 2019;Orlando et al, 2017b;Li et al, 2018a,b;Liu et al, 2018;Pal et al, 2018). Sensitivity and specificity (Chen et al, 2015b;Christopher et al, 2018;Fu et al, 2018;Gómez-Valverde et al, 2019;Li et al, 2018a;Liu et al, 2018) are also used in different studies to complement the AUC when targetting binary classification outcomes. Accuracy was reported in (Cerentinia et al, 2018;Raghavendra et al, 2018) as another evaluation metric, although this metric might be biased if the proportion of non-glaucomatous images is significantly higher than the glaucomatous ones (Orlando et al, 2017a).…”

A new platform designed for glaucoma screening: identifying the risk of glaucomatous optic neuropathy using fundus photography with deep learning architecture together with intraocular pressure measurements

“…Fu et al reported sensitivity 0.85 and specificity 0.84 [29]. In a review paper by Gómez-Valverde et al the performances of a number of different convolutional neural networks were presented [20]. As a result, the sensitivity in the range 0.79-0.92, and the specificity in the range 0.75-0.91 were achieved.…”

“…Thus, there is a need for development of automatic methods for GON detection based on fundus images. Several reports have proved the efficacy of machine learning in glaucoma [9][10][11][12][13][14][15][16][17][18][19][20][21][22][23].…”

Deep learning for objective OCTA detection of diabetic retinopathy

“…Transfer learning is a training method to adopt some weights of a pre-trained CNN and appropriately re-train the CNN to optimize the weights for a specific task, i.e., AI classification of retinal images [31]. In fundus photography, transfer learning has been explored to conduct artery-vein segmentation [32], glaucoma detection [33,34], and diabetic macular thinning assessment [35]. Recently, transfer learning has also been explored in OCT for detecting choroidal neovascularization (CNV) and diabetic macular edema (DME) [31], and AMD [36].…”