RESUMO
PURPOSE: To develop deep learning (DL) algorithm to detect glaucoma progression using optical coherence tomography (OCT) images, in the absence of a reference standard. DESIGN: Retrospective cohort study. METHODS: Glaucomatous and healthy eyes with ≥5 reliable peripapillary OCT (Spectralis, Heidelberg Engineering) circle scans were included. A weakly supervised time-series learning model, called noise positive-unlabeled (Noise-PU) DL was developed to classify whether sequences of OCT B-scans showed glaucoma progression. The model used 2 learning schemes, one to identify age-related changes by differentiating test sequences from glaucoma vs healthy eyes, and the other to identify test-retest variability based on scrambled OCTs of glaucoma eyes. Both models' bases were convolutional neural networks (CNN) and long short-term memory (LSTM) networks which were combined to form a CNN-LSTM model. Model features were combined and jointly trained to identify glaucoma progression, accounting for age-related loss. The DL model's outcomes were compared with ordinary least squares (OLS) regression of retinal nerve fiber layer (RNFL) thickness over time, matched for specificity. The hit ratio was used as a proxy for sensitivity. RESULTS: Eight thousand seven hundred eighty-five follow-up sequences of 5 consecutive OCT tests from 3253 eyes (1859 subjects) were included in the study. The mean follow-up time was 3.5 ± 1.6 years. In the test sample, the hit ratios of the DL and OLS methods were 0.498 (95%CI: 0.470-0.526) and 0.284 (95%CI: 0.258-0.309) respectively (P < .001) when the specificities were equalized to 95%. CONCLUSION: A DL model was able to identify longitudinal glaucomatous structural changes in OCT B-scans using a surrogate reference standard for progression.
RESUMO
PURPOSE: To evaluate the performance of an intensive, clustered testing approach in identifying eyes with rapid glaucoma progression over 6 months in the Fast Progression Assessment through Clustered Evaluation (Fast-PACE) Study. DESIGN: Prospective cohort study. PARTICIPANTS: A total of 125 eyes from 65 primary open-angle glaucoma (POAG) subjects. METHODS: Subjects underwent 2 sets of 5 weekly visits (clusters) separated by an average of 6 months and then were followed with single visits every 6 months for an overall mean follow-up of 25 months (mean of 17 tests). Each visit consisted of testing with standard automated perimetry (SAP) 24-2 and 10-2, and spectral-domain OCT (SD-OCT). Progression was assessed using trend analyses of SAP mean deviation (MD) and retinal nerve fiber layer (RNFL) thickness. Generalized estimating equations were applied to adjust for correlations between eyes for confidence interval (CI) estimation and hypothesis testing. MAIN OUTCOME MEASURES: Diagnostic accuracy of the 6-month clustering period to identify progression detected during the overall follow-up. RESULTS: A total of 19 of 125 eyes (15%, CI, 9%-24%) progressed based on SAP 24-2 MD over the 6-month clustering period. A total of 14 eyes (11%, CI, 6%-20%) progressed on SAP 10-2 MD, and 16 eyes (13%, CI, 8%-21%) progressed by RNFL thickness, with 30 of 125 eyes (24%, CI, 16%-34%) progressing by function, structure, or both. Of the 35 eyes progressing during the overall follow-up, 25 had progressed during the 6-month clustering period, for a sensitivity of 71% (CI, 53%-85%). Of the 90 eyes that did not progress during the overall follow-up, 85 also did not progress during the 6-month period, for a specificity of 94% (CI, 88%-98%). Of the 14 eyes considered fast progressors by SAP 24-2, SAP 10-2, or SD-OCT during the overall follow-up, 13 were identified as progressing during the 6-month cluster period, for a sensitivity of 93% (CI, 66%-100%) for identifying fast progression with a specificity of 85% (CI, 77%-90%). CONCLUSIONS: Clustered testing in the Fast-PACE Study detected fast-progressing glaucoma eyes over 6 months. The methodology could be applied in clinical trials investigating interventions to slow glaucoma progression and may be of value for short-term assessment of high-risk subjects. FINANCIAL DISCLOSURE(S): Proprietary or commercial disclosure may be found after the references in the Footnotes and Disclosures at the end of this article.
