Human Versus Machine: Comparing a Deep Learning Algorithm to Human Gradings for Detecting Glaucoma on Fundus Photographs.

To compare the diagnostic performance of human gradings vs predictions provided by a machine-to-machine (M2M) deep learning (DL) algorithm trained to quantify retinal nerve fiber layer (RNFL) damage on fundus photographs. Evaluation of a machine learning algorithm. An M2M DL algorithm trained with RNFL thickness parameters from spectral-domain optical coherence tomography was applied to a subset of 490 fundus photos of 490 eyes of 370 subjects graded by 2 glaucoma specialists for the probability of glaucomatous optical neuropathy (GON), and estimates of cup-to-disc (C/D) ratios. Spearman correlations with standard automated perimetry (SAP) global indices were compared between the human gradings vs the M2M DL-predicted RNFL thickness values. The area under the receiver operating characteristic curves (AUC) and partial AUC for the region of clinically meaningful specificity (85%-100%) were used to compare the ability of each output to discriminate eyes with repeatable glaucomatous SAP defects vs eyes with normal fields. The M2M DL-predicted RNFL thickness had a significantly stronger absolute correlation with SAP mean deviation (rho=0.54) than the probability of GON given by human graders (rho=0.48; P < .001). The partial AUC for the M2M DL algorithm was significantly higher than that for the probability of GON by human graders (partial AUC = 0.529 vs 0.411, respectively; P = .016). An M2M DL algorithm performed as well as, if not better than, human graders at detecting eyes with repeatable glaucomatous visual field loss. This DL algorithm could potentially replace human graders in population screening efforts for glaucoma.

Jammal AA ，Thompson AC ，Mariottoni EB ，Berchuck SI ，Urata CN ，Estrela T ，Wakil SM ，Costa VP ，Medeiros FA ... -  《-》

From Machine to Machine: An OCT-Trained Deep Learning Algorithm for Objective Quantification of Glaucomatous Damage in Fundus Photographs.

Previous approaches using deep learning (DL) algorithms to classify glaucomatous damage on fundus photographs have been limited by the requirement for human labeling of a reference training set. We propose a new approach using quantitative spectral-domain (SD) OCT data to train a DL algorithm to quantify glaucomatous structural damage on optic disc photographs. Cross-sectional study. A total of 32 820 pairs of optic disc photographs and SD OCT retinal nerve fiber layer (RNFL) scans from 2312 eyes of 1198 participants. The sample was divided randomly into validation plus training (80%) and test (20%) sets, with randomization performed at the patient level. A DL convolutional neural network was trained to assess optic disc photographs and predict SD OCT average RNFL thickness. The DL algorithm performance was evaluated in the test sample by evaluating correlation and agreement between the predictions and actual SD OCT measurements. We also assessed the ability to discriminate eyes with glaucomatous visual field loss from healthy eyes with area under the receiver operating characteristic (ROC) curves. The mean prediction of average RNFL thickness from all 6292 optic disc photographs in the test set was 83.3±14.5 μm, whereas the mean average RNFL thickness from all corresponding SD OCT scans was 82.5±16.8 μm (P = 0.164). There was a very strong correlation between predicted and observed RNFL thickness values (Pearson r = 0.832; R2 = 69.3%; P < 0.001), with mean absolute error of the predictions of 7.39 μm. The area under the ROC curves for discriminating glaucomatous from healthy eyes with the DL predictions and actual SD OCT average RNFL thickness measurements were 0.944 (95% confidence interval [CI], 0.912-0.966) and 0.940 (95% CI, 0.902-0.966), respectively (P = 0.724). We introduced a novel DL approach to assess fundus photographs and provide quantitative information about the amount of neural damage that can be used to diagnose and stage glaucoma. In addition, training neural networks to predict SD OCT data objectively represents a new approach that overcomes limitations of human labeling and could be useful in other areas of ophthalmology.

Medeiros FA ，Jammal AA ，Thompson AC 《-》

A Deep Learning Algorithm to Quantify Neuroretinal Rim Loss From Optic Disc Photographs.

To train a deep learning (DL) algorithm that quantifies glaucomatous neuroretinal damage on fundus photographs using the minimum rim width relative to Bruch membrane opening (BMO-MRW) from spectral-domain optical coherence tomography (SDOCT). Cross-sectional study. A total of 9282 pairs of optic disc photographs and SDOCT optic nerve head scans from 927 eyes of 490 subjects were randomly divided into the validation plus training (80%) and test sets (20%). A DL convolutional neural network was trained to predict the SDOCT BMO-MRW global and sector values when evaluating optic disc photographs. The predictions of the DL network were compared to the actual SDOCT measurements. The area under the receiver operating curve (AUC) was used to evaluate the ability of the network to discriminate glaucomatous visual field loss from normal eyes. The DL predictions of global BMO-MRW from all optic disc photographs in the test set (mean ± standard deviation [SD]: 228.8 ± 63.1 μm) were highly correlated with the observed values from SDOCT (mean ± SD: 226.0 ± 73.8 μm) (Pearson's r = 0.88; R2 = 77%; P < .001), with mean absolute error of the predictions of 27.8 μm. The AUCs for discriminating glaucomatous from healthy eyes with the DL predictions and actual SDOCT global BMO-MRW measurements were 0.945 (95% confidence interval [CI]: 0.874-0.980) and 0.933 (95% CI: 0.856-0.975), respectively (P = .587). A DL network can be trained to quantify the amount of neuroretinal damage on optic disc photographs using SDOCT BMO-MRW as a reference. This algorithm showed high accuracy for glaucoma detection, and may potentially eliminate the need for human gradings of disc photographs.

Thompson AC ，Jammal AA ，Medeiros FA 《-》

Assessment of a Segmentation-Free Deep Learning Algorithm for Diagnosing Glaucoma From Optical Coherence Tomography Scans.

Thompson AC ，Jammal AA ，Berchuck SI ，Mariottoni EB ，Medeiros FA ... -  《-》

Detection of Progressive Glaucomatous Optic Nerve Damage on Fundus Photographs with Deep Learning.

To investigate whether predictions of retinal nerve fiber layer (RNFL) thickness obtained from a deep learning model applied to fundus photographs can detect progressive glaucomatous changes over time. Retrospective cohort study. Eighty-six thousand one hundred twenty-three pairs of color fundus photographs and spectral-domain (SD) OCT images collected during 21 232 visits from 8831 eyes of 5529 patients with glaucoma or glaucoma suspects. A deep learning convolutional neural network was trained to assess fundus photographs and to predict SD OCT global RNFL thickness measurements. The model then was tested on an independent sample of eyes that had longitudinal follow-up with both fundus photography and SD OCT. The ability to detect eyes that had statistically significant slopes of SD OCT change was assessed by receiver operating characteristic (ROC) curves. The repeatability of RNFL thickness predictions was investigated by measurements obtained from multiple photographs that had been acquired during the same day. The relationship between change in predicted RNFL thickness from photographs and change in SD OCT RNFL thickness over time. The test sample consisted of 33 466 pairs of fundus photographs and SD OCT images collected during 7125 visits from 1147 eyes of 717 patients. Eyes in the test sample were followed up for an average of 5.3 ± 3.3 years, with an average of 6.2 ± 3.8 visits. A significant correlation was found between change over time in predicted and observed RNFL thickness (r = 0.76; 95% confidence interval [CI], 0.70-0.80; P < 0.001). Retinal nerve fiber layer predictions showed an ROC curve area of 0.86 (95% CI, 0.83-0.88) to discriminate progressors from nonprogressors. For detecting fast progressors (slope faster than 2 μm/year), the ROC curve area was 0.96 (95% CI, 0.94-0.98), with a sensitivity of 97% for 80% specificity and 85% for 90% specificity. For photographs obtained at the same visit, the intraclass correlation coefficient was 0.946 (95% CI, 0.940-0.952), with a coefficient of variation of 3.2% (95% CI, 3.1%-3.3%). A deep learning model was able to obtain objective and quantitative estimates of RNFL thickness that correlated well with SD OCT measurements and potentially could be used to monitor for glaucomatous changes over time.

Medeiros FA ，Jammal AA ，Mariottoni EB 《-》