Deep learning based retinal OCT segmentation.

We look at the recent application of deep learning (DL) methods in automated fine-grained segmentation of spectral domain optical coherence tomography (OCT) images of the retina. We describe a new method combining fully convolutional networks (FCN) with Gaussian Processes for post processing. We report performance comparisons between the proposed approach, human clinicians, and other machine learning (ML) such as graph based approaches. The approach is demonstrated on an OCT dataset consisting of mild non-proliferative diabetic retinopathy from the University of Miami. The method is shown to have performance on par with humans, also compares favorably with the other ML methods, and appears to have as small or smaller mean unsigned error (equal to 1.06), versus errors ranging from 1.17 to 1.81 for other methods, and compared with human error of 1.10.

Observational study of subclinical diabetic macular edema.

PURPOSE: To determine the rate of progression of eyes with subclinical diabetic macular edema (DME) to clinically apparent DME or DME necessitating treatment during a 2-year period. METHODS: In all, 43 eyes from 39 study participants with subclinical DME, defined as absence of foveal center edema as determined with slit lamp biomicroscopy but a center point thickness (CPT) between 225 and 299 µm on time domain (Stratus, Carl Zeiss Meditec) optical coherence tomography (OCT) scan, were enrolled from 891 eyes of 582 subjects screened. Eyes were evaluated annually for up to 2 years for the primary outcome, which was an increase in OCT CPT of at least 50 µm from baseline and a CPT of at least 300 µm, or treatment for DME (performed at the discretion of the investigator). RESULTS: The cumulative probability of meeting an increase in OCT CPT of at least 50 µm from baseline and a CPT of at least 300 µm, or treatment for DME was 27% (95% confidence interval (CI): 14%, 38%) by 1 year and 38% (95% CI: 23%, 50%) by 2 years. CONCLUSIONS: Although subclinical DME may be uncommon, this study suggests that between approximately one-quarter and one-half of eyes with subclinical DME will progress to more definite thickening or be judged to need treatment for DME within 2 years after its identification.

Photodynamic therapy with verteporfin for choroidal neovascularization in patients with diabetic retinopathy.

PURPOSE: To report the use of photodynamic therapy (PDT) with verteporfin in three patients with choroidal neovascularization (CNV) from age-related macular degeneration and underlying diabetic retinopathy. The level of diabetic retinopathy would have excluded these patients from participation in previously reported randomized clinical trials evaluating PDT with verteporfin due to a theoretic concern of damage to the overlying retinal vasculature. DESIGN: Retrospective interventional case series. METHODS: Three patients from a referral practice with at least severe nonproliferative diabetic retinopathy and a history of clinically significant macular edema developed loss of vision from concurrent choroidal neovascularization evaluated with fundus photography and fluorescein angiography before and after PDT with verteporfin to identify adverse retinal vascular events. RESULTS: Four eyes in three patients had PDT using verteporfin. Three eyes received two treatments. With short follow-up, visual acuity remained stable in two eyes, improved from 20/400 to 20/320 in one eye, and decreased from 20/200 to 20/400 in one eye. Fluorescein angiograms at intervals from 2 weeks to 3 months after PDT showed no damage to the retinal vasculature or progression of the diabetic retinopathy, but did show a decreased area of fluorescein leakage from CNV. One eye that had new subretinal hemorrhage following treatment appeared to show new vasculopathy on initial evaluation of the post-treatment angiogram. Retrospective review suggested that the subretinal hemorrhage provided increased contrast to more easily visualize vasculopathy that was present before the PDT. CONCLUSIONS: Three patients with diabetic retinopathy undergoing a total of seven PDT treatments with verteporfin in four eyes had no new retinal vascular abnormalities develop. No other atypical responses of CNV to PDT were noted except new subretinal hemorrhage, providing increased contrast of the overlying vasculature, which gave the false impression of the development of new vasculopathy in one eye. Patients with diabetic retinopathy who have concurrent CNV for which PDT with verteporfin is recommended should be cautioned regarding the theoretical concerns of harming the retinal vasculature. Periodic surveillance for such concerns seems warranted until more experience is obtained.

Photodynamic therapy of subfoveal choroidal neovascularization in age-related macular degeneration with verteporfin: two-year results of 2 randomized clinical trials-tap report 2.

OBJECTIVE: To report 24-month vision and fluorescein angiographic outcomes from trials evaluating photodynamic therapy with verteporfin (Visudyne; CIBA Vision Corp, Duluth, Ga) in patients with subfoveal choroidal neovascularization (CNV) caused by age-related macular degeneration (AMD). DESIGN: Two multicenter, double-masked, placebo-controlled, randomized clinical trials. SETTING: Twenty-two ophthalmology practices in Europe and North America. PARTICIPANTS: Patients with subfoveal CNV lesions caused by AMD with greatest linear dimension on the retina measuring 5400 micrometer or less, with evidence of classic CNV and best-corrected visual acuity (approximate Snellen equivalent) between 20/40 and 20/200. METHODS: The methods were similar to those described in our 1-year results, with follow-up examinations beyond 1 year continuing every 3 months (except for Photograph Reading Center evaluations, which occurred only at month 18 and month 24 examinations). During the second year, the same regimen (with verteporfin or placebo as applied at baseline) was used if angiography showed fluorescein leakage from CNV. The primary outcome was the proportion of eyes with fewer than 15 letters (approximately 3 lines) of visual acuity loss at the month 24 examination, adhering to an intent-to-treat analysis. The last observation was carried forward to impute for any missing data. RESULTS: Three hundred fifty-one (87%) of 402 patients in the verteporfin group compared with 178 (86%) of 207 patients in the placebo group completed the month 24 examination. Beneficial outcomes with respect to visual acuity and contrast sensitivity noted at the month 12 examination in verteporfin-treated patients were sustained through the month 24 examination. At the month 24 examination for the primary outcome, 213 (53%) of 402 verteporfin-treated patients compared with 78 (38%) of 207 placebo-treated patients lost fewer than 15 letters (P<.001). In subgroup analyses for predominantly classic lesions (in which the area of classic CNV makes up at least 50% of the area of the entire lesion) at baseline, 94 (59%) of 159 verteporfin-treated patients compared with 26 (31%) of 83 placebo-treated patients lost fewer than 15 letters at the month 24 examination (P<.001). For minimally classic lesions (in which the area of classic CNV makes up <50% but >0% of the area of the entire lesion) at baseline, no statistically significant differences in visual acuity were noted. Few additional photosensitivity adverse reactions and injection site adverse events were associated with verteporfin therapy in the second year of follow-up. CONCLUSIONS: The visual acuity benefits of verteporfin therapy for AMD patients with predominantly classic CNV subfoveal lesions are safely sustained for 2 years, providing more compelling evidence to use verteporfin therapy for these cases. For AMD patients with subfoveal lesions that are minimally classic, there is insufficient evidence to warrant routine use of verteporfin therapy.

Manifest refraction versus autorefraction for patients with subfoveal choroidal neovascularization.

PURPOSE: To compare the results from manifest refraction using trial lenses and a standard visual acuity protocol to results from autorefraction for obtaining refractive error and best corrected visual acuity in patients enrolled in a randomized clinical trial. METHODS: During a 4-month period, 29 patients with subfoveal choroidal neovascularization (CNV), who were enrolled in the Submacular Surgery Trials (SSTs) Pilot Study at the Wilmer Ophthalmological Institute, gave verbal consent to participate in this study. Best corrected visual acuity was obtained using Early Treatment Diabetic Retinopathy Study (ETDRS) visual acuity charts and standardized room lighting after performance of manifest refraction, according to the SST protocol, and autorefraction. Refractive error (spherical equivalent) and visual acuity scores were obtained in both eyes of all patients. RESULTS: On average, manifest refraction gave a spherical equivalent that was 1.04 D more plus than autorefraction (95% limits of agreement = 0.74, 1.34). On average, the visual acuity score was 1.5 letters better after manifest refraction than after autorefraction (95% limits of agreement = 0, 3.0). The comparison of the two methods of refraction was subdivided according to visual acuity level and eye disease (age-related macular degeneration or ocular histoplasmosis syndrome). CONCLUSIONS: Despite large differences in spherical equivalent between manifest refraction and autorefraction, the visual acuity scores were close (mean difference, 1.5 letters). Other studies comparing subjective refraction and autorefraction have shown similar results. Autorefraction in patients with subfoveal CNV may be a satisfactory alternative to manifest refraction in clinical trials and field studies in which best corrected visual acuity is of interest.

Adult-onset foveomacular pigment epithelial dystrophy: clinicopathologic correlation of three cases.

PURPOSE: The authors describe the clinicopathologic features of three patients with adult onset foveomacular pigment epithelial dystrophy (AOFPED). METHODS: The eyes of three patients were studied ophthalmoscopically and by fluorescein angiography, and obtained postmortem and studied by light and electron microscopy. RESULTS: Histopathologic study of the three patient's eyes disclosed central loss of the retinal pigment epithelium and photoreceptor cell layer with a moderate number of pigment-containing macrophages present in the subretinal space and outer retina. To either side, the retinal pigment epithelium was distended with much lipofuscin. Basal laminar and basal linear deposits were present throughout the central area. No discontinuities of Bruch membrane were present. CONCLUSION: The findings in the eyes of three patients with AOFPED included marked aging changes that are similar to those seen in age-related macular degeneration. Pigmented cells with lipofuscin in the subretinal space account for the vitelliform appearance.

Submacular surgery trials randomized pilot trial of laser photocoagulation versus surgery for recurrent choroidal neovascularization secondary to age-related macular degeneration: I. Ophthalmic outcomes submacular surgery trials pilot study report number 1.

PURPOSE: To report complications and changes in vision during 2 years of follow-up of patients with age-related macular degeneration assigned randomly to surgical removal or to laser photocoagulation of subfoveal recurrent neovascular lesions in a pilot trial designed to test methods, to refine estimates of outcome rates, and to project patient accrual rates for a larger multicenter randomized trial to evaluate submacular surgery. PATIENTS AND METHODS: Eligible patients with previous laser photocoagulation of extrafoveal or juxtafoveal choroidal neovascularization secondary to age-related macular degeneration were enrolled at 15 collaborating clinical centers. Assignments to treatment arm were made by personnel at a central coordinating center. Adherence to eligibility criteria and treatment assignment was assessed centrally at a photograph reading center. Patients were examined at 3, 6, 12, and 24 months after treatment for data collection purposes. Outcome measures reported include treatment complications, adverse events, requirements for additional treatment, and 2-year changes in visual acuity from baseline. RESULTS: Of 70 patients enrolled, 36 were assigned to laser photocoagulation and 34 to submacular surgery; all were treated as assigned. One patient in each group died before the 2-year examination. Visual acuity was measured at the 2-year examination for 31 of the surviving patients (89%) in the laser arm and for 28 of the surviving patients (85%) in the surgery arm. The 2-year measurements for 36 of the 59 patients (61%) were made by an examiner masked to treatment assignment and to the identity of the study eye. Improvements and losses of visual acuity were observed in both treatment arms; 20 of 31 study eyes (65%) in the laser arm and 14 of 28 study eyes (50%) in the surgery arm had visual acuity 2 years after enrollment that was better than or no more than 1 line worse than the baseline level. Changes in visual acuity and the size of the central macular lesions from baseline to the 2-year examination were similar in the treatment arms. Few serious complications were observed in either arm at the time of initial treatment; serious adverse events were rare. During follow-up, 11 laser-treated eyes and 18 surgically treated eyes had additional intraocular procedures. CONCLUSIONS: The data from this pilot trial suggest no reason to prefer submacular surgery over laser photocoagulation for treatment of patients with age-related macular degeneration who have lesions similar to those studied in this pilot trial. Any clinical trial designed to compare submacular surgery with laser photocoagulation in eyes with age-related macular degeneration and subfoveal recurrent neovascular lesions must enroll several hundred patients in order to reach a statistically valid conclusion regarding differences between these two methods of treatment with respect to either changes in visual acuity or complication rates.

A preliminary study of photodynamic therapy using verteporfin for choroidal neovascularization in pathologic myopia, ocular histoplasmosis syndrome, angioid streaks, and idiopathic causes.

OBJECTIVE: To evaluate short-term safety and the effects on visual acuity and fluorescein angiography of single or multiple sessions of photodynamic therapy with verteporfin for choroidal neovascularization (CNV) not related to age-related macular degeneration (AMD), including pathologic myopia, the ocular histoplasmosis syndrome, angioid streaks, and idiopathic causes. DESIGN: A nonrandomized, multicenter, open-label, dose-escalation phase 1 and 2 clinical trial. SETTING: Four ophthalmic centers in Europe and North America providing retinal care. PARTICIPANTS: Thirteen patients with subfoveal CNV due to pathologic myopia, the ocular histoplasmosis syndrome, angioid streaks, or idiopathic causes. METHODS: Standardized protocol refraction, visual acuity testing, ophthalmic examinations, color photographs, and fluorescein angiograms were used to evaluate the results of photodynamic therapy treatments with verteporfin. Follow-up ranged from 12 weeks for patients who were treated once to 43 weeks for patients who were treated up to 4 times. RESULTS: Verteporfin therapy was well tolerated in patients with CNV not related to AMD. No deterioration in visual acuity was observed; most patients gained at least 1 line of vision. Reduction in the size of leakage area from classic CNV was noted in all patients as early as 1 week after verteporfin therapy, with complete absence of leakage from classic CNV in almost half of the patients. Improvement in visual acuity after verteporfin therapy was greatest (+6, +8, and +9 lines) in 3 patients with relatively poor initial visual acuity (between 20/200 and 20/800). Up to 4 treatments were found to have short-term safety even with retreatment intervals as short as 4 weeks. CONCLUSIONS: Treatment of CNV not related to AMD with verteporfin therapy achieves short-term cessation of fluorescein leakage from CNV in a small number of patients without loss of vision. Further randomized clinical trials including a larger number of patients are under way to confirm whether verteporfin therapy is beneficial for subfoveal CNV not related to AMD.

Ketorolac treatment of pseudophakic cystoid macular edema identified more than 24 months after cataract extraction.

OBJECTIVE: To evaluate whether ketorolac ophthalmic drops prescribed four times a day can be associated with improved visual acuity and prompt resolution of edema for patients with pseudophakic cystoid macular edema identified more than 24 months after cataract surgery. DESIGN: Prospective, nonrandomized, comparative (subject self-controlled) trial. PARTICIPANTS: The records of nine patients who had pseudophakic cystoid macular edema more than 24 months after cataract surgery at the time treatment commenced were identified at the Wilmer Retinal Vascular Center from September 1, 1996, through March 1, 1997. MAIN OUTCOME MEASURES: Best-corrected visual acuities measured on a retroilluminated Bailey-Lovie chart approximately every 3 months, contact lens biomicroscopy, and fluorescein angiography following ketorolac. INTERVENTION: Commercially available ketorolac ophthalmic drops 0.5% were prescribed for the affected eye four times a day for at least 3 months and continued until edema resolved. RESULTS: Ten eyes of nine patients were identified more than 24 months after cataract extraction (median, 59 months). Seven eyes (70%) improved (mean, +3.2 lines; range, +1 to +13 lines), including six by 2 or more lines 3 months after treatment initiation. Two eyes (20%) were unchanged, and one eye (10%) was 1 line worse. All seven eyes that improved 1 line or more had some or complete angiographic resolution of fluorescein dye leakage. In these seven eyes, ketorolac was discontinued when dye leakage completely resolved or failed to continue to improve on periodic 3-month follow-up examinations. In all seven eyes, recurrence of edema was noted within 3 months after ketorolac was stopped. CONCLUSIONS: Chronic pseudophakic cystoid macular edema identified more than 24 months after cataract surgery can improve with topical ketorolac but probably requires persistent use.

Enlargement of atrophy and visual acuity loss in the geographic atrophy form of age-related macular degeneration.

OBJECTIVE: To describe the progression of geographic atrophy (GA) from age-related macular degeneration (AMD) with respect to visual acuity (VA) loss and enlargement of atrophy. DESIGN: A prospectively observed case series. SETTING: Tertiary retinal referral center. PARTICIPANTS: One hundred twenty-three patients with GA due to AMD who completed at least 1 year of follow-up (median follow-up, 3 years) were examined annually. METHODS: At each examination, a protocol best-corrected VA of each eye was measured, a clinical examination was performed, and color fundus photographs were taken. The areas of atrophy were drawn and measured. MAIN OUTCOME MEASURES: Visual acuity loss and enlargement of total and central atrophy. RESULTS: At baseline, median VA was poorer with larger areas of atrophy, but there was wide variation related to sparing of the fovea. Thirty-one percent of all study eyes suffered a three-line VA loss from baseline by 2 years, and 53% had a three-line loss by 4 years. Those eyes with VA better than 20/50 had the highest rate of acuity loss; 27% of these eyes had acuities of 20/200 or worse at 4 years. Visual acuity loss in the GA study eye was similar in patients with bilateral GA and in those with choroidal neovascularization in the fellow eye. Total atrophy enlarged a median of 1.8 Macular Photocoagulation Study disc areas (DA) at 2 years; atrophy within a 4-DA circle centered on the fovea enlarged a median of 0.9 DA. Two (22%) of nine patients with GA in one eye and only drusen without advanced AMD in the fellow eye developed GA in the fellow eye at 2 years. CONCLUSIONS: Geographic atrophy is associated with a significant decline in VA over time in many eyes. Areas of atrophy continue to enlarge over time, even when already large at baseline. The combination of reduced VA with enlargement of atrophy, occurring bilaterally in most patients, can lead to significant impairment of visual function.

Photodynamic therapy with verteporfin for choroidal neovascularization caused by age-related macular degeneration: results of a single treatment in a phase 1 and 2 study.

OBJECTIVE: To evaluate the safety and short-term visual and fluorescein angiographic effects of a single photodynamic therapy treatment with verteporfin with the use of different dosage regimens in patients with choroidal neovascularization (CNV) from age-related macular degeneration. DESIGN: Nonrandomized, multicenter, open-label, clinical trial using 5 dosage regimens. SETTING: Four ophthalmic centers in North America and Europe providing retinal care. PARTICIPANTS: Patients with subfoveal CNV caused by age-related macular degeneration. METHODS: Standardized protocol refraction, visual acuity testing, ophthalmic examination, color photographs, and fluorescein angiograms were used to evaluate the effects of a single treatment of photodynamic therapy with verteporfin. Follow-up was planned through 3 months in 97 patients and for less than 3 months in 31 other patients. RESULTS: The mean visual acuity change (and range of change) from baseline at the follow-up examination at week 12 after a single treatment with regimens 1 through 5 was -0.2 (-3 to +2), -0.9 (-9 to +5), -1.6 (-9 to +2), +0.4 (-8 to +7), and +0.1 (-8 to +9) lines, respectively. Only the highest light dose (150 J/cm2) in regimens 2 and 3, which produced angiographic nonperfusion of neurosensory retinal vessels, caused marked vision loss. Some cessation of fluorescein leakage from CNV was achieved without loss of vision when the light dose used was less than 150 J/cm2. Systemic adverse events were rare. Cessation of fluorescein leakage from CNV was noted in all regimens by 1 week after photodynamic therapy. Fluorescein leakage from at least a portion of the CNV reappeared by 4 to 12 weeks after treatment in almost all cases. Progression of classic CNV beyond the area of CNV identified before treatment was noted in 42 (51%) of the 83 eyes with classic CNV followed up for 3 months after a single treatment. Eyes in which the area of any CNV leakage at 12 weeks was less than at baseline had a significantly better visual acuity outcome (+0.8 line) than eyes in which CNV leakage progressed (-0.8 line). CONCLUSIONS: Photodynamic therapy with verteporfin achieved short-term cessation of fluorescein leakage from CNV without loss of vision or growth of classic CNV in some patients with age-related macular degeneration. Except for nonperfusion of neurosensory retinal vessels at a light dose of 150 J/cm2, no other adverse events were of concern. Randomized clinical trials to investigate whether this new modality can preserve vision in patients with CNV secondary to age-related macular degeneration are justified.

Photodynamic therapy with verteporfin for choroidal neovascularization caused by age-related macular degeneration: results of retreatments in a phase 1 and 2 study.

OBJECTIVES: To evaluate safety and short-term visual acuity and fluorescein angiographic effects of photodynamic therapy (PDT) after retreatments with verteporfin for choroidal neovascularization (CNV) in age-related macular degeneration (AMD) that demonstrated fluorescein leakage after at least 1 course of PDT. DESIGN: Nonrandomized, multicenter, open-label phase 1 and 2 clinical trial using 2 different retreatment dosage regimens. SETTING: Four ophthalmic centers in Europe and North America providing retinal care. METHODS: Standardized protocol refraction, visual acuity testing, ophthalmic examinations, color photographs, and fluorescein angiograms were used to evaluate the results of multiple PDT treatments. Two regimens (regimens 2 and 4) for treatment and retreatment were chosen from 5 used in a single-treatment study. Both regimens used a verteporfin dose of 6 mg/m2 infused for 10 minutes. However, regimen 2 used a light dose of 100 J/cm2 applied 20 minutes after the start of the verteporfin infusion, whereas regimen 4 used a light dose of 50, 75, or 100 J/cm2 applied 15 minutes after infusion commenced. Posttreatment evaluations were planned in 31 participants up to 3 months after up to 2 retreatments given at 2- or 4-week intervals after initial PDT treatment. Similar posttreatment evaluations were planned after retreatments in 5 additional participants who were reenrolled some time more than 12 weeks after an initial PDT treatment. RESULTS: The average visual acuity change for the 31 participants who had retreatment within 2 to 4 weeks after the initial treatment and a follow-up examination 16 to 20 weeks after the initial treatment was 0.2 lines (range, -4 to 4 lines) in regimen 2 and -1.0 line (range, -5 to 3 lines) in regimen 4. Similar outcomes were noted in the 5 reenrolled participants. Cessation of fluorescein leakage from classic CNV for at least 1 to 4 weeks could be achieved without loss of visual acuity after at least 2 treatments in 2 (6.5%) of 31 patients. Similar to single-treatment effects, the disappearance of leakage was documented regularly at 1 week after each retreatment. Fluorescein leakage reappeared by 4 to 12 weeks after a retreatment in almost all cases. However, compared with baseline, leakage activity appeared to be reduced after multiple PDT courses. For the 31 patients who had follow-up for 3 months after the last retreatment and had received retreatment 2 to 4 weeks after the initial treatment, progression of CNV beyond the area identified before the retreatment was noted in 10 (48%) of the 21 eyes with classic CNV in regimen 2 and 9 (90%) of 10 eyes in regimen 4. The rate and severity of ocular or systemic adverse events were not increased by multiple applications. CONCLUSIONS: Multiple applications of PDT with verteporfin achieve repetitive, short-term cessation of fluorescein leakage from CNV secondary to AMD, without loss of visual acuity. This strategy can be used in randomized clinical trials investigating the efficacy of verteporfin in PDT for recurrent fluorescein dye leakage from persistent or recurrent CNV, following an initial or subsequent PDT treatment, with maintenance of visual acuity. Retreatments may achieve progressive cessation of leakage and prevent further growth of CNV and subsequent visual loss.