Abstract
Purpose :
Currently, no Convolutional Neural Network (CNN)-based program for diagnosis of glaucoma has been developed. We designed this cross-sectional study to investigate the performance of CNN to identify glaucomatous visual fields (VFs) from non-glaucomatous VFs and to compare the performance of machine against human ophthalmologists.
Methods :
In this cross-sectional study, 4012 VF tests obtained by Humphrey Field Analyzer from 3 different ophthalmic centers in mainland China were collected. Reliability criteria were established as fixation losses less than 2/13, false positive and false negative rates lower than 15%. All the VFs from both eyes of a single patient are assigned to either train or validation set to avoid data leakage. In this way, we split a total of 3917 Pattern Deviation images from 1352 patients into two sets, 3617 for training and 300 for validation. We exploit CNN with VGG architecture to classify the VFs from glaucoma and non-glaucoma patients. Accuracy of diagnosis by human ophthalmologists, traditional rules (AGIS and GSS2), traditional machine learning algorithms and CNN were compared.
Results :
We totally collected 4012 VF reports, including glaucoma and non-glaucoma reports. There is no significant difference between left to right ratio (P = 0.6211), while age (P = 0.0022), VFI (P = 0.0001), MD (P = 0.0039) and PSD (P = 0.0001) exhibited obvious statistical significance. On the validation set of 300 VFs, our algorithm based on CNN achieved an accuracy of 0.876, while the specificity and sensitivity are 0.826 and 0.932, respectively. For ophthalmologists, the average accuracies are 0.607, 0.585 and 0.626 for resident ophthalmologists, attending ophthalmologists and glaucoma experts, respectively. AGIS and GSS2 achieved accuracy of 0.459 and 0.523 respectively. Three traditional machine learning algorithms, namely support vector machine (SVM), random forest (RF), and k-nearest neighbor (k-NN) were also implemented and evaluated in the experiments, which achieved accuracy of 0.670, 0.644, and 0.591 respectively.
Conclusions :
Our algorithm based on CNN has achieved higher accuracy compared to human ophthalmologists and traditional rules (AGIS and GSS2). It will be a powerful tool to distinguish glaucoma from non-glaucoma VFs, and may help screening and diagnosis of glaucoma in the future.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.