AI-Human Hybrid Workflow Enhances Teleophthalmology for the Detection of Diabetic Retinopathy.

Objective: Detection of diabetic retinopathy (DR) outside of specialized eye care settings is an important means of access to vision-preserving health maintenance. Remote interpretation of fundus photographs acquired in a primary care or other nonophthalmic setting in a store-and-forward manner is a predominant paradigm of teleophthalmology screening programs. Artificial intelligence (AI)-based image interpretation offers an alternative means of DR detection. IDx-DR (Digital Diagnostics Inc) is a Food and Drug Administration-authorized autonomous testing device for DR. We evaluated the diagnostic performance of IDx-DR compared with human-based teleophthalmology over 2 and a half years. Additionally, we evaluated an AI-human hybrid workflow that combines AI-system evaluation with human expert-based assessment for referable cases. Design: Prospective cohort study and retrospective analysis. Participants: Diabetic patients ≥ 18 years old without a prior DR diagnosis or DR examination in the past year presenting for routine DR screening in a primary care clinic. Methods: Macula-centered and optic nerve-centered fundus photographs were evaluated by an AI algorithm followed by consensus-based overreading by retina specialists at the Stanford Ophthalmic Reading Center. Detection of more-than-mild diabetic retinopathy (MTMDR) was compared with in-person examination by a retina specialist. Main Outcome Measures: Sensitivity, specificity, accuracy, positive predictive value, and gradability achieved by the AI algorithm and retina specialists. Results: The AI algorithm had higher sensitivity (95.5% sensitivity; 95% confidence interval [CI], 86.7%-100%) but lower specificity (60.3% specificity; 95% CI, 47.7%-72.9%) for detection of MTMDR compared with remote image interpretation by retina specialists (69.5% sensitivity; 95% CI, 50.7%-88.3%; 96.9% specificity; 95% CI, 93.5%-100%). Gradability of encounters was also lower for the AI algorithm (62.5%) compared with retina specialists (93.1%). A 2-step AI-human hybrid workflow in which the AI algorithm initially rendered an assessment followed by overread by a retina specialist of MTMDR-positive encounters resulted in a sensitivity of 95.5% (95% CI, 86.7%-100%) and a specificity of 98.2% (95% CI, 94.6%-100%). Similarly, a 2-step overread by retina specialists of AI-ungradable encounters improved gradability from 63.5% to 95.6% of encounters. Conclusions: Implementation of an AI-human hybrid teleophthalmology workflow may both decrease reliance on human specialist effort and improve diagnostic accuracy. Financial Disclosures: Proprietary or commercial disclosure may be found after the references.

Displacement of submacular hemorrhage with intravitreal tissue plasminogen activator following 27 gauge transvitreal fine needle aspiration biopsy for choroidal melanoma.

PURPOSE: To describe the management of submacular hemorrhage (SMH), a vision threatening complication following transvitreal choroidal biopsy, with intravitreal tissue plasminogen activator (tPA) and pure perfluoropropane (C3F8) gas bubble injection. OBSERVATIONS: A 53 year old female with choroidal melanoma of the left eye underwent iodine-125 plaque brachytherapy placement and 27 gauge transvitreal fine needle aspiration choroidal biopsy for gene expression profiling. On postoperative day 2, large SMH was identified on dilated fundus examination. At the time of plaque brachytherapy removal, intravitreal tPA and pure C3F8 gas bubble injection with post operative positioning was also performed to attempt displacement of SMH. At postoperative month 1 following tPA and gas bubble displacement, the SMH was completely displaced inferotemporally outside of the macula and visual acuity improved from 20/70 at postoperative week 1 to 20/25 at postoperative month 1. CONCLUSIONS AND IMPORTANCE: Subretinal hemorrhage can be a complication of transvitreal choroidal tumor biopsy but early detection and prompt treatment can result in good visual outcomes.

"Giant cell arteritis manifesting as retinal arterial occlusion and paracentral acute middle maculopathy in a patient on pembrolizumab for metastatic uveal melanoma".

PURPOSE: To report the association of pembrolizumab, an immune checkpoint inhibitor (ICI), with giant cell arteritis (GCA) presenting as paracentral acute middle maculopathy (PAMM) secondary to retinal arterial occlusion. OBSERVATIONS: 86-year old male with history of treated choroidal melanoma now with metastatic uveal melanoma to the liver on pembrolizumab, an ICI, who presented with acute vision loss in the uninvolved left eye. Spectral domain optical coherence tomography showed band-like increased hyperreflectivity in the middle retinal layers at the level of the inner nuclear layer consistent with PAMM. Intravenous fluorescein angiogram demonstrated significant delay in filling of the superotemporal and inferotemporal arteries with nonperfusion of the temporal retina consistent with multiple branch retinal arterial occlusions. Work-up for GCA was performed and temporal artery biopsy showed healed arteritis. CONCLUSIONS AND IMPORTANCE: Pembrolizumab can cause ocular and life-threatening systemic adverse effects and as use of ICIs has increased, it is important to be aware of these associations. There should be a low threshold for GCA work up in patients on ICI therapy who present with acute vision loss and evidence of retinal occlusive disease with or without classic GCA systemic symptoms.

Argus II Retinal Prosthesis Implantation in a Patient With a History Of Strabismus Surgery:A Surgical Technique.

BACKGROUND AND OBJECTIVE: The authors describe a modified surgical technique for Argus II retinal prosthesis implantation. PATIENTS AND METHODS: The modified surgical technique was performed in a 43-year-old male with retinitis pigmentosa (RP) who underwent strabismus surgery in that eye at age 8. RESULTS: During Argus II implantation, the lateral rectus (LR) was noted to be 10 mm posterior to its original insertion due to prior strabismus surgery. The authors placed three 5-0 nylon mattress sutures at the expected location of the LR to support the external portion of the prosthesis, specifically the coil. The band was then passed through the mattress sutures, underneath the inferior, and medial recti with temporal placement of the coil. CONCLUSION: Placement of three mattress sutures allowed proper positioning of the device and enabled fixation of the suture tabs, thus minimizing the risk of postoperative conjunctival erosion and device exposure. [Ophthalmic Surg Lasers Imaging Retina. 2019;50:116-119.].

OUTCOMES OF REPEAT PARS PLANA VITRECTOMY AFTER FAILED SURGERY FOR PROLIFERATIVE VITREORETINOPATHY.

PURPOSE: To evaluate outcomes of repeat pars plana vitrectomy for proliferative vitreoretinopathy after previous failed pars plana vitrectomy. METHODS: This is a retrospective case series including 51 eyes of 50 patients who underwent repeat surgery after failed previous pars plana vitrectomy for proliferative vitreoretinopathy from 2000 to 2015 at the Kresge Eye Institute, Detroit, MI. Patients were classified into successful and unsuccessful groups. Success was defined as retinal reattachment, silicone oil removed, and best-corrected visual acuity (BCVA) ≥5/200 at the final follow-up visit. RESULTS: Forty-three eyes (84.3%) were successfully reattached at the last follow-up. Seventeen (33.3%) eyes were deemed successful and 34 (66.7%) eyes unsuccessful according to our criteria. Compared with the successful group, eyes in the unsuccessful group had more eyes with preoperative BCVA <5/200 (P < 0.001), preoperative BCVA of hand motion or worse (P = 0.002), preoperative flare ≥Grade 2+ (P = 0.03), preoperative posterior breaks (P = 0.02), previous retinectomy (P = 0.04), and final postoperative hypotony (intraocular pressure ≤ 5 mmHg) (P = 0.005). Eyes with silicone oil removed were more likely to have BCVA ≥5/200 (P < 0.001) at the final follow-up visit. Location of patients >100 miles (P = 0.04) from Detroit and preoperative BCVA of hand motion or worse (P = 0.01) were significantly associated with failure in the logistic regression analysis. CONCLUSION: Success after repeat surgery for proliferative vitreoretinopathy should include ambulatory vision, retinal reattachment, and silicone oil removal. We identified several preoperative and perioperative factors that were associated with success in the bivariate and logistic analyses. The decision to perform surgical reoperation in these patients should be based on multiple factors, most importantly preoperative BCVA.

LOSS OF EXTERNAL LIMITING MEMBRANE INTEGRITY PREDICTS PROGRESSION OF HYDROXYCHLOROQUINE RETINAL TOXICITY AFTER DRUG DISCONTINUATION.

PURPOSE: To quantify external limiting membrane (ELM) disruption and photoreceptor volume over time, using spectral domain optical coherence tomography in eyes with hydroxychloroquine (HCQ) toxic effects after discontinuation. METHODS: We performed a retrospective chart review of patients who were screened for toxic effects of HCQ between January 1, 2009 and August 31, 2014, and identified 10 patients diagnosed as having HCQ retinal toxic effects. Intact ELM and the Bruch membrane were manually traced using ImageJ software and their lengths from each scan of the spectral domain optical coherence tomography macular volume were summed. The ratio of intact ELM length/Bruch membrane length was calculated. We measured the two-dimensional photoreceptor area between the intact ELM and Bruch membrane in every cross-sectional B-scan. We calculated the total volume of photoreceptors in a volumetric OCT by multiplying photoreceptor area by the distance between B-scans. RESULTS: Of the 10 patients (120 eyes), 1 eye was excluded because of the presence of vitreomacular traction. The mean cumulative dose of HCQ was 1,951 g (range, 584-3,650 g). The mean follow-up duration was 34.1 months (range, 11-64 months). Based on the extent of ELM integrity at diagnosis of HCQ toxicity, we classified eyes into 2 groups: 1) 12 eyes showed severe ELM disruption and 2) 7 eyes had intact ELM. All 19 eyes showed characteristic signs of HCQ toxicity on the visual field examination at the baseline. Seven of 12 eyes with ELM disruption showed progressive ELM changes, including 7 eyes that additionally developed progressive photoreceptor volume decline, whereas 5 eyes remained stable. All 7 eyes with intact ELM remained stable over time. CONCLUSION: Intact ELM status at the time of HCQ discontinuation is a good prognostic sign, and ELM status may allow clinicians to predict patients at risk for progression of retinopathy. Quantitative measurements of ELM disruption and photoreceptors volume may provide an objective tool to monitor outer retinal changes due to HCQ.

LONGITUDINAL QUANTITATIVE EVALUATION OF PHOTORECEPTOR VOLUME FOLLOWING REPAIR OF MACULA-OFF RETINAL DETACHMENT.

PURPOSE: To quantify photoreceptor volume changes after successful surgical repair of macula-off retinal detachment and to correlate these volumetric changes to postoperative best-corrected visual acuity (BCVA). METHODS: Retrospective study of 15 eyes of 15 patients with macula-off retinal detachment who underwent successful surgical repair. A minimum of 4 optical coherence tomography scans that straddled the foveal center was used to quantify the central photoreceptor volume (central 1 mm). RESULTS: Mean photoreceptor volume at the first postoperative visit was 0.451 mm, increasing to 0.523 mm at the final postoperative visit (P = 0.004). Mean BCVA improved from 1.13 ± 0.59 logarithm of the minimum angle of resolution units (∼20/270) preoperatively to 0.52 ± 0.42 logarithm of the minimum angle of resolution units (∼20/66) at the final postoperative visit (P = 0.001). Mean photoreceptor volume at either the initial or final visit demonstrated significant correlations with final postoperative BCVA (r = -0.670, P = 0.017 and r = -0.753, P = 0.005, respectively). Shorter time interval from diagnosis to surgery was significantly associated with greater mean final postoperative photoreceptor volume (r = -0.588, P = 0.021) and better mean final postoperative BCVA (r = 0.709, P = 0.003). CONCLUSION: We observed a significant increase in photoreceptor volume after successful retinal detachment repair; photoreceptor volume was positively associated with BCVA and time to surgery. Our series emphasizes the importance of prompt surgical repair and shows that photoreceptor recovery and volumetric improvement correlate significantly with BCVA.

Effect of anti-VEGF treatment on choroidal thickness over time in patients with neovascular age-related macular degeneration.

PURPOSE: To evaluate change in subfoveal choroidal thickness (SCT) as measured by spectral-domain optical coherence tomography (SD-OCT) in patients with neovascular age-related macular degeneration (NVAMD) undergoing anti-vascular endothelial growth factor (VEGF) therapy. METHODS: Patients with a diagnosis of NVAMD were retrospectively reviewed to identify those who had at least 12 months of follow-up. The SCT was manually measured from Bruch membrane to the choroid-sclera junction at baseline and last follow-up. Only cases in which the choroid was fully visible were included in quantitative analyses. The SCT measurements were correlated with other characteristics including number and duration of treatments. RESULTS: Sixty eyes of 47 patients with a follow-up of 23.8 months (SD 7.3) met study inclusion criteria, and 49 eyes of 40 patients received anti-VEGF treatment. Mean age was 83.7 years, and 52% were female. Treated eyes received a mean of 7.8 (SD 7.3) intravitreal anti-VEGF injections. The SCT at baseline was 126.7 µm (SD 50.6) for untreated and 136.2 µm (SD 57.6) for treated eyes. The SCT showed a decrease over time in both groups, with a mean rate of reduction of 6.0 µm (p<0.0002) in treated eyes and 3.6 µm (p = 0.3741) in untreated eyes. However, the change in SCT did not differ between the groups (p = 0.5113), and did not correlate with the number of re-treatments (p = 0.552), visual acuity at baseline (p = 0.618), or change in visual acuity over time (p = 0.429). CONCLUSIONS: Although choroidal thickness decreased over time in eyes with NVAMD, anti-VEGF therapy did not appear to accelerate or otherwise alter this decline.

Structural and functional implications of severe foveal dystopia in epiretinal membranes.

PURPOSE: The purpose of this study was to examine the functional and structural correlates of severe foveal dystopia in patients with epiretinal membranes. METHODS: For this retrospective study of 29 eyes with epiretinal membrane, we identified 7 eyes that had severe foveal dystopia (defined as fovea located >200 µm from its expected location) and followed the direction and rate of foveal movement pre- and postoperatively. RESULTS: Epiretinal membrane traction caused the fovea to move preoperatively at a rate of 275 µm/month from its anatomical location in 2 patients. The final preoperative foveal location was, on average, 1,217 ± 683 µm away from its expected location. Postoperatively, foveal movement toward its expected location was largest during the first month after surgery (mean = 547 ± 340 µm) and slowed down until the final follow-up position was achieved (mean = 301 ± 131 µm). Overall, the fovea moved a total of 848 ± 445 µm, allowing the fovea to correct only 32.8 ± 22.1% of the total displacement from its expected location. A univariate regression model confirmed a linear relationship between preoperative visual acuity and preoperative foveal distance from its expected anatomical location with an R of 0.759 (P = 0.0107). CONCLUSION: The extent of tractional foveal dystopia correlates with decreased visual acuity. Although all patients experienced functional and anatomical improvements with surgery, long-standing or severe foveal dystopia may be associated with permanent structural changes that limit functional outcome. Cases with extreme degrees of foveal dystopia may benefit from early intervention to prevent irreversible structural and functional changes.

Comparison of manually corrected retinal thickness measurements from multiple spectral-domain optical coherence tomography instruments.

BACKGROUND/AIMS: To compare retinal thickness measurements from three different spectral domain optical coherence instruments when manual segmentation is employed to standardise retinal boundary locations. METHODS: 40 eyes of 21 healthy subjects were scanned on the Cirrus HD-OCT, Topcon 3D-OCT-2000 and Heidelberg Spectralis-OCT. Raw data were imported into custom grading software (3D-OCTOR). Manual segmentation was performed on every data set, and retinal thickness values in the foveal central subfield were computed. RESULTS: 37 eyes of 20 subjects were gradable on every machine. The average retinal thicknesses for these eyes were 236.7 µm (SD 20.1), 235.7 µm (SD 20.4) and 236.5 µm (SD 18.0) for the Cirrus, 3D-OCT-2000 and Spectralis, respectively. Comparing manual retinal thickness measurements between any two machines, the maximum difference was 18.2 µm. The mean absolute differences per eye between two machines were: 4.9 µm for Cirrus versus 3D-OCT-2000, 3.7 µm for Cirrus versus Spectralis and 4.4 µm for 3D-OCT-2000 versus Spectralis. CONCLUSIONS: When a uniform position is used to locate the outer retinal boundary, the retinal thickness measurements derived from three different spectral domain optical coherence instruments devices are virtually identical. Manual correction may allow OCT-derived thickness measurements to be compared between devices in clinical trials and clinical research.