Estimation of Visual Function Using Deep Learning From Ultra-Widefield Fundus Images of Eyes With Retinitis Pigmentosa.

Importance: There is no widespread effective treatment to halt the progression of retinitis pigmentosa. Consequently, adequate assessment and estimation of residual visual function are important clinically. Objective: To examine whether deep learning can accurately estimate the visual function of patients with retinitis pigmentosa by using ultra-widefield fundus images obtained on concurrent visits. Design, Setting, and Participants: Data for this multicenter, retrospective, cross-sectional study were collected between January 1, 2012, and December 31, 2018. This study included 695 consecutive patients with retinitis pigmentosa who were examined at 5 institutions. Each of the 3 types of input images-ultra-widefield pseudocolor images, ultra-widefield fundus autofluorescence images, and both ultra-widefield pseudocolor and fundus autofluorescence images-was paired with 1 of the 31 types of ensemble models constructed from 5 deep learning models (Visual Geometry Group-16, Residual Network-50, InceptionV3, DenseNet121, and EfficientNetB0). We used 848, 212, and 214 images for the training, validation, and testing data, respectively. All data from 1 institution were used for the independent testing data. Data analysis was performed from June 7, 2021, to December 5, 2022. Main Outcomes and Measures: The mean deviation on the Humphrey field analyzer, central retinal sensitivity, and best-corrected visual acuity were estimated. The image type-ensemble model combination that yielded the smallest mean absolute error was defined as the model with the best estimation accuracy. After removal of the bias of including both eyes with the generalized linear mixed model, correlations between the actual values of the testing data and the estimated values by the best accuracy model were examined by calculating standardized regression coefficients and P values. Results: The study included 1274 eyes of 695 patients. A total of 385 patients were female (55.4%), and the mean (SD) age was 53.9 (17.2) years. Among the 3 types of images, the model using ultra-widefield fundus autofluorescence images alone provided the best estimation accuracy for mean deviation, central sensitivity, and visual acuity. Standardized regression coefficients were 0.684 (95% CI, 0.567-0.802) for the mean deviation estimation, 0.697 (95% CI, 0.590-0.804) for the central sensitivity estimation, and 0.309 (95% CI, 0.187-0.430) for the visual acuity estimation (all P < .001). Conclusions and Relevance: Results of this study suggest that the visual function estimation in patients with retinitis pigmentosa from ultra-widefield fundus autofluorescence images using deep learning might help assess disease progression objectively. Findings also suggest that deep learning models might monitor the progression of retinitis pigmentosa efficiently during follow-up.

Comparative outcomes of trabeculotomy ab externo versus trabecular ablation ab interno for open angle glaucoma.

PURPOSE: To compare clinical outcomes between trabeculotomy ab externo (TLO) and trabecular ablation ab interno using a Trabectome (TAT) for open angle glaucoma (OAG). STUDY DESIGN: Prospective and retrospective cohort study. METHODS: We prospectively recruited 68 patients who underwent TAT between May 2013 and July 2015, and retrospectively recruited 59 patients who underwent TLO between April 2010 and May 2013. We included OAG patients who could be followed up for more than 12 months after surgery. Surgical outcomes were compared between groups. Kaplan-Meier analyses and Cox proportional hazards' models were used for four sets of criteria using postoperative intraocular pressures (IOP) ≤ 18 or ≤ 16 mmHg, and/or showing ≥ 20% reduction from baseline. RESULTS: Mean preoperative IOPs were 22.6 ± 7.4 mmHg in the TAT group and 24.3 ± 6.6 mmHg in the TLO group. Mean postoperative IOPs in the TAT group were 15.8 ± 3.6 mmHg at 12-months, and 15.7 ± 5.5 mmHg at 36-months, and 16.0 ± 3.4 mmHg at 12 months and 15.2 ± 3.8 mmHg at 36 months in the TLO group. Postoperatively, success rates at 36-months were significantly higher in the TLO group in some criteria; IOP reductions at 1- and 2- years were significantly higher in the TLO group; these differences were not significant after adjustment with other potential prognostic factors. The number of glaucoma medications were significantly higher in the TAT group, at 2-years (P < 0.001) but not at 3-years (P = 0.22). CONCLUSION: Surgical successes following TLO and TAT are not significantly different 3 years postoperatively after adjustment of potential prognostic factors.

Retinal Blood Flow Velocity Change in Parafoveal Capillary after Topical Tafluprost Treatment in Eyes with Primary Open-angle Glaucoma.

Although ocular circulation at the retina and optic disc is known to be associated with the pathology of glaucoma, direct measurement of blood flow velocity has been difficult to obtain. This prospective observational study enrolled 11 consecutive patients with treatment-naïve primary open-angle glaucoma (POAG) and 11 healthy subjects, and the effects of topical tafluprost treatment on ocular circulation were examined at baseline and at 1, 4, and 12 weeks after initiating treatment with topical tafluprost on POAG patients using multiple modalities, which include adaptive optics scanning laser ophthalmoscopy (AOSLO). Baseline mean intraocular pressure (IOP) was significantly higher and mean parafoveal blood flow velocity (pBFV) was significantly lower in POAG eyes than in healthy eyes. Mean IOP was significantly decreased (1 week, -19.1%; 4 weeks, -17.7%; and 12 weeks, -23.5%; all P < 0.001) and mean pBFV was significantly increased from the baseline at all follow-up periods after initiating treatment (1 week, 14.9%, P = 0.007; 4 weeks, 21.3%, P < 0.001; and 12 weeks, 14.3%, P = 0.002). These results reveal that tafluprost may not only lower IOP but may also improve retinal circulation in POAG eyes and AOSLO may be useful to evaluate retinal circulatory change after treatment.

Alterations in the neural and connective tissue components of glaucomatous cupping after glaucoma surgery using swept-source optical coherence tomography.

PURPOSE: To visualize changes in deep optic nerve head (ONH) structures following glaucoma surgery using (3-dimensional [3D]) swept-source optical coherence tomography (SS-OCT) and to determine the clinical and structural factors associated with postoperative lamina cribrosa (LC) and prelaminar neural tissue (PLT) changes. METHODS: In this prospective observational case series, SS-OCT thin-sliced datasets of the ONH covering a 3- × 3-mm area comprised of 256 B-scans (interval between scans = ∼12 µm) were obtained before and 3 months after the surgery and evaluated in 73 eyes of 73 patients with glaucoma. Bruch's membrane opening (BMO) and anterior LC boundary were manually delineated by two methods; one in every four B-scans (64 B-scans per eye) and 15 equally spaced horizontal B-scans in BMO area, excluding both ends (interval between scans = 96-120 µm). After former delineation, the point with maximum LC depth among 64 B-scans was automatically calculated, and LC depth and PLT thickness were averaged among 5 points adding 4 points 100 µm apart from this point vertically and horizontally. Associations between the percent change in LC depth and other clinical and structural parameters were tested for statistical analysis. RESULTS: Lamina cribrosa depth and axial length significantly decreased and PLT thickness significantly increased after surgery. The percent change of maximum LC depth correlated significantly with the percent change of IOP (P = 0.008), baseline LC depth (P = 0.032), and visual field mean deviation (P = 0.035; at the point with maximum LC depth), while the percent change of axial length correlated with IOP reduction (P = 0.002) but not with visual field mean deviation. CONCLUSIONS: Swept-source optical coherence tomography enables 3D analysis of deep ONH structures, and the change in LC depth after glaucoma surgery have association with IOP change and the severity of glaucomatous optic neuropathy.

Detection of localized retinal nerve fiber layer defects in glaucoma using enhanced spectral-domain optical coherence tomography.

OBJECTIVE: To compare retinal nerve fiber layer (RNFL) defects on fundus photographs with circumpapillary RNFL (cpRNFL) thinning or disruption on images obtained by speckle-noise-reduced spectral-domain optical coherence tomography (enhanced SD OCT), single-scan SD OCT, and single-scan time-domain OCT (TD OCT). DESIGN: Retrospective, comparative case series. PARTICIPANTS: Forty-four eyes of 44 patients with open-angle glaucoma with localized, wedge-shaped RNFL defects on red-free photographs and 35 normal eyes of 35 volunteers. METHODS: Cross-sectional images of the cpRNFL and cpRNFL thinning, compared with the confidence interval limit of the normative database where the RNFL defect was photographically identified, were compared between the 3 types of OCT instruments: enhanced SD OCT (SD OCT with eye tracking and averaging of 16 images at the same location to reduce speckle noise; Spectralis HRA+OCT; Heidelberg Engineering, Heidelberg, Germany), single-scan SD OCT (RTVue-100; Optovue, Fremont, CA), and single-scan TD OCT (Stratus; Carl Zeiss-Meditec, Dublin, CA). MAIN OUTCOME MEASURES: Cross-sectional images of localized RNFL defects on red-free fundus photographs, sensitivity for detecting the photographic RNFL defect, and sensitivity and specificity for detecting glaucoma as having at least 1 abnormally thinned sector on the cpRNFL thickness map on OCT. RESULTS: Among the 44 eyes with glaucoma, 65 RNFL defects were identified on red-free fundus photographs. The cpRNFL boundaries were clearer on enhanced SD OCT images than on single-scan SD OCT or TD OCT images, particularly in regions corresponding to the RNFL defects. Enhanced SD OCT revealed various degrees of cpRNFL thinning, and disruption of cpRNFL reflectivity was seen in the same location as the photographic RNFL defect for 23 (35.4%) of the 65 RNFL defects. The RNFL defects were significantly less likely to be detected by single-scan TD OCT or SD OCT (P = 0.002 and P = 0.006, respectively) when the RNFL was not disrupted. Enhanced SD OCT was more sensitive in detecting the RNFL defects that were not disrupted compared with single-scan TD OCT (P<0.0001) or SD OCT (P<0.0001). Enhanced SD OCT had better sensitivity and specificity for detecting glaucoma compared with single-scan TD OCT or SD OCT (sensitivity, P = 0.006 and P = 0.001; specificity, P = 0.001 and P = 0.004, respectively). CONCLUSIONS: These results suggest that speckle-noise reduction can improve the detection of photographic RNFL defects in which cpRNFL reflectivity on OCT images is not disrupted. FINANCIAL DISCLOSURE(S): Proprietary or commercial disclosure may be found after the references.