A deep learning model for the detection of both advanced and early glaucoma using fundus photography

Abstract

PurposeTo build a deep learning model to diagnose glaucoma using fundus photography.DesignCross sectional case study Subjects, Participants and Controls: A total of 1,542 photos (786 normal controls, 467 advanced glaucoma and 289 early glaucoma patients) were obtained by fundus photography.MethodThe whole dataset of 1,542 images were split into 754 training, 324 validation and 464 test datasets. These datasets were used to construct simple logistic classification and convolutional neural network using Tensorflow. The same datasets were used to fine tune pre-trained GoogleNet Inception v3 model.ResultsThe simple logistic classification model showed a training accuracy of 82.9%, validation accuracy of 79.9% and test accuracy of 77.2%. Convolutional neural network achieved accuracy and area under the receiver operating characteristic curve (AUROC) of 92.2% and 0.98 on the training data, 88.6% and 0.95 on the validation data, and 87.9% and 0.94 on the test data. Transfer-learned GoogleNet Inception v3 model achieved accuracy and AUROC of 99.7% and 0.99 on training data, 87.7% and 0.95 on validation data, and 84.5% and 0.93 on test data.ConclusionBoth advanced and early glaucoma could be correctly detected via machine learning, using only fundus photographs. Our new model that is trained using convolutional neural network is more efficient for the diagnosis of early glaucoma than previously published models.

Accurate prediction of glaucoma from colour fundus images with a convolutional neural network that relies on active and transfer learning

“…The merits of transfer learning in the field of automated glaucoma detection using fundus images have been illustrated using both small (Ahn et al. ; Shibata et al. ) and large (Christopher et al.…”

Automatic detection of rare pathologies in fundus photographs using few-shot learning

“…Email address: gwenole.quellec@inserm.fr (Gwenolé Quellec) 2019; Ahn et al, 2019;Diaz-Pinto et al, 2019) and age-related macular degeneration (AMD) (Matsuba et al, 2018;Pead et al, 2019), the other two major sight-threatening pathologies in developed countries. Other pathologies such as retinopathy of prematurity (Wang et al, 2018) have also been targeted.…”

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

“…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].…”

“…Recent studies of DR using deep machine learning showed high sensitivity and specificity for the detection of changes typical for DR [11,12,15]. The detection of glaucoma may be more challenging than the diagnosis of DR because it relies on estimation of subtle changes in the ONH shape and cupping including the stage of glaucoma and refractive errors of the eye [9,14].…”