Prediction of retinopathy progression using deep learning on retinal images within the Scottish screening programme.

BACKGROUND/AIMS: National guidelines of many countries set screening intervals for diabetic retinopathy (DR) based on grading of the last screening retinal images. We explore the potential of deep learning (DL) on images to predict progression to referable DR beyond DR grading, and the potential impact on assigned screening intervals, within the Scottish screening programme. METHODS: We consider 21 346 and 247 233 people with type 1 diabetes mellitus (T1DM) and type 2 diabetes mellitus (T2DM), respectively, each contributing on average 4.8 and 4.4 screening intervals of which 1339 and 4675 intervals concluded with a referable screening episode. Information extracted from fundus images using DL was used to predict referable status at the end of interval and its predictive value in comparison to screening-assigned DR grade was assessed. RESULTS: The DL predictor increased the area under the receiver operating characteristic curve in comparison to a predictor using current DR grades from 0.809 to 0.87 for T1DM and from 0.825 to 0.87 for T2DM. Expected sojourn time-the time from becoming referable to being rescreened-was found to be 3.4 (T1DM) and 2.7 (T2DM) weeks less for a DL-derived policy compared with the current recall policy. CONCLUSIONS: We showed that, compared with using the current retinopathy grade, DL of fundus images significantly improves the prediction of incident referable retinopathy before the next screening episode. This can impact screening recall interval policy positively, for example, by reducing the expected time with referable disease for a fixed workload-which we show as an exemplar. Additionally, it could be used to optimise workload for a fixed sojourn time.

Deep learning detection of diabetic retinopathy in Scotland's diabetic eye screening programme.

BACKGROUND/AIMS: Support vector machine-based automated grading (known as iGradingM) has been shown to be safe, cost-effective and robust in the diabetic retinopathy (DR) screening (DES) programme in Scotland. It triages screening episodes as gradable with no DR versus manual grading required. The study aim was to develop a deep learning-based autograder using images and gradings from DES and to compare its performance with that of iGradingM. METHODS: Retinal images, quality assurance (QA) data and routine DR grades were obtained from national datasets in 179 944 patients for years 2006-2016. QA grades were available for 744 images. We developed a deep learning-based algorithm to detect whether either eye contained ungradable images or any DR. The sensitivity and specificity were evaluated against consensus QA grades and routine grades. RESULTS: Images used in QA which were ungradable or with DR were detected by deep learning with better specificity compared with manual graders (p<0.001) and with iGradingM (p<0.001) at the same sensitivities. Any DR according to the DES final grade was detected with 89.19% (270 392/303 154) sensitivity and 77.41% (500 945/647 158) specificity. Observable disease and referable disease were detected with sensitivities of 96.58% (16 613/17 201) and 98.48% (22 600/22 948), respectively. Overall, 43.84% of screening episodes would require manual grading. CONCLUSION: A deep learning-based system for DR grading was evaluated in QA data and images from 11 years in 50% of people attending a national DR screening programme. The system could reduce the manual grading workload at the same sensitivity compared with the current automated grading system.

Can deep learning on retinal images augment known risk factors for cardiovascular disease prediction in diabetes? A prospective cohort study from the national screening programme in Scotland.

AIMS: This study's objective was to evaluate whether deep learning (DL) on retinal photographs from a diabetic retinopathy screening programme improve prediction of incident cardiovascular disease (CVD). METHODS: DL models were trained to jointly predict future CVD risk and CVD risk factors and used to output a DL score. Poisson regression models including clinical risk factors with and without a DL score were fitted to study cohorts with 2,072 and 38,730 incident CVD events in type 1 (T1DM) and type 2 diabetes (T2DM) respectively. RESULTS: DL scores were independently associated with incident CVD with adjusted standardised incidence rate ratios of 1.14 (P = 3 × 10-04 95 % CI (1.06, 1.23)) and 1.16 (P = 4 × 10-33 95 % CI (1.13, 1.18)) in T1DM and T2DM cohorts respectively. The differences in predictive performance between models with and without a DL score were statistically significant (differences in test log-likelihood 6.7 and 51.1 natural log units) but the increments in C-statistics from 0.820 to 0.822 and from 0.709 to 0.711 for T1DM and T2DM respectively, were small. CONCLUSIONS: These results show that in people with diabetes, retinal photographs contain information on future CVD risk. However for this to contribute appreciably to clinical prediction of CVD further approaches, including exploitation of serial images, need to be evaluated.

RETINAL VASCULAR BED AREA IN EYES WITH RETINAL VEIN OCCLUSION ON ULTRA-WIDEFIELD FLUORESCEIN ANGIOGRAPHY: WAVE Study.

PURPOSE: To compute retinal vascular bed area (RVBA) in square millimeters on distortion corrected ultra-widefield fluorescein angiography images in eyes with retinal vein occlusion (RVO). METHODS: Prospective observational study. The peripheral distortion of baseline ultra-widefield fluorescein angiography (Optos 200Tx) images of 30 patients with RVO from the WAVE study (NCT01710839) and 13 control eyes of normal subjects was corrected using the stereographic projection method to compute RVBA in square millimeters. RESULTS: In comparison with age- and sex-matched normal control eyes, eyes with RVO had a decreased global RVBA for the entire retina (50.5 ± 20.4 mm 2 vs. 62.6 ± 12.2 mm 2 , P = 0.023). Eyes with RVO and the unaffected fellow eye had a similar RVBA globally (50.5 ± 20.4 mm 2 vs. 46.2 ± 18.9 mm 2 , P = 0.523). The RVBA was observed to negatively correlate with nonperfusion area (R = -0.47, P = 0.009). However, RVBA was not related to the severity of macular edema ( P > 0.05). CONCLUSION: Eyes with RVO have a similar RVBA to the unaffected fellow eyes but with a reduction when compared with normal control eyes. Retinal vascular bed area appears to be a surrogate biomarker of retinal ischemia on ultra-widefield fluorescein angiography but not the extent of macular edema.

Association of Maximizing Visible Retinal Area by Manual Eyelid Lifting With Grading of Diabetic Retinopathy Severity and Detection of Predominantly Peripheral Lesions When Using Ultra-Widefield Imaging.

IMPORTANCE: Methods that increase visible retinal area (VRA; measured in millimeters squared) may improve identification of diabetic retinopathy (DR) lesions. OBJECTIVE: To evaluate the association of dilation and manual eyelid lifting (MLL) with VRA on ultra-widefield imaging (UWFI) and the association of VRA with grading of DR severity and detection of predominantly peripheral lesions (PPLs). DESIGN, SETTING, AND PARTICIPANTS: Retrospective, comparative case-control study at the Joslin Diabetes Center, Boston, Massachusetts. Nonmydriatic UWFI with MLL was acquired from a DR teleophthalmology program (Joslin Vision Network [JVN]). A second cohort of mydriatic UWFI was acquired at an academic retina practice (Beetham Eye Institute [BEI]) from November 6, 2017, to November 6, 2018, and with MLL thereafter until November 6, 2019. Fully automated algorithms determined VRA and hemorrhage and/or microaneurysm (HMA) counts. Predominantly peripheral lesions and HMAs were defined as present when at least 1 field had greater HMA number in the peripheral retina than within the corresponding Early Treatment Diabetic Retinopathy Study field. Participants included 3014 consecutive patients (5919 eyes) undergoing retinal imaging at JVN and BEI. EXPOSURES: Dilation and MLL performed at the time of UWFI. MAIN OUTCOMES AND MEASURES: Visible retinal area, DR severity, and presence of PPLs. RESULTS: Of the 3014 participants, mean (SD) age was 56.1 (14.5) years, 1302 (43.2%) were female, 2450 (81.3%) were White, and mean (SD) diabetes duration was 15.9 (11.4) years. All images from 5919 eyes with UWFI were analyzed. Mean (SD) VRA was 665.1 (167.6) mm2 for all eyes (theoretical maximal VRA, 923.9 mm2), 550.8 (240.7) mm2 for nonmydriatic JVN with MLL (1418 eyes [24.0%]), 688.1 (119.9) mm2 for mydriatic BEI images (3650 eyes [61.7%]), and 757.0 (69.7) mm2 for mydriatic and MLL BEI images (851 eyes [14.4%]). Dilation increased VRA by 25% (P < .001) and MLL increased VRA an additional 10% (P < .001). Nonmydriatic MLL increased VRA by 11.0%. With MLL, HMA counts in UWFI fields increased by 41.7% (from 4.8 to 6.8; P < .001). Visible retinal area was moderately associated with increasing PPL-HMA overall and in each cohort (all, r = 0.33; BEI, r = 0.29; JVN, r = 0.36; P < .001). In JVN images, increasing VRA was associated with more PPL-HMA (quartile 1 [Q1], 23.7%; Q2, 45.8%; Q3, 60.6%; and Q4, 69.2%; P < .001). CONCLUSIONS AND RELEVANCE: Using fully automated VRA and HMA detection algorithms, pupillary dilation and eyelid lifting were shown to substantially increase VRA and PLL-HMA detection. Given the importance of HMA and PPL for determining risk of DR progression, these findings emphasize the importance of maximizing VRA for optimal risk assessment in clinical trials and teleophthalmology programs.

Retinal vascular bed area on ultra-wide field fluorescein angiography indicates the severity of diabetic retinopathy.

AIMS: To quantify retinal vascular bed area (RVBA) in square millimetres on stereographically projected ultra-wide field (UWF) fluorescein angiography (FA) in eyes with diabetic retinopathy (DR). METHODS: A prospective, observational study. Baseline Optos 200Tx UWF FA images of 80 eyes with DR from the DAVE (NCT01552408) and RECOVERY (NCT02863354) studies were stereographically projected at the Doheny Image Reading Center to adjust for peripheral distortion. The early-phase FA frame was used to extract the retinal vasculature as a mask for calculating RVBA. The pixels of the retinal vasculature were automatically computed in square millimetres using manufacturer-provided software. RESULTS: Eighteen of 80 diabetic eyes were excluded because image quality and contrast were insufficient for automatic extraction of the retinal vasculature from the background fluorescence. The remaining 62 eyes were included in the final analysis. In comparison with age-matched and sex-matched normal controls, eyes with DR had a higher global RVBA for the entire retina (p<0.001), and RVBA correlated with DR severity (p<0.001), with a higher RVBA in eyes with proliferative DR (66.1±16.2 mm2) than in those with non-proliferative DR (56.2±16.6 mm2) or in normal controls (37.2±9.9 mm2). This tendency was also present in the posterior retina and mid-periphery but absent in the far-periphery. RVBA did not correlate with retinal ischaemia (p>0.05). CONCLUSIONS: Eyes with DR harboured a larger global RVBA for the entire retina than normal controls, and RVBA appeared to indicate DR severity. However, this biomarker was not observed to be a good indicator of retinal ischaemia.

NEW BIOMARKER QUANTIFYING THE EFFECT OF ANTI-VEGF THERAPY IN EYES WITH PROLIFERATIVE DIABETIC RETINOPATHY ON ULTRAWIDE FIELD FLUORESCEIN ANGIOGRAPHY: RECOVERY Study.

PURPOSE: To quantify changes of the retinal vascular bed area (RVBA) in mm2 on stereographically projected ultrawide field fluorescein angiography images in eyes with proliferative diabetic retinopathy after antivascular endothelial growth factor injection. METHODS: This is a prospective, observational study. The early-phase ultrawide field fluorescein angiography images (Optos 200Tx) of 40 eyes with proliferative diabetic retinopathy and significant nonperfusion obtained at baseline and after six months (NCT02863354) were stereographically projected by correcting peripheral distortion. The global retinal vasculature on ultrawide field fluorescein angiography was extracted for calculating RVBA by summing the real size (mm2) of all the pixels automatically. RESULTS: For the entire cohort, the global RVBA for the entire retina decreased from 67.1 ± 15.5 to 43.6 ± 18.8 mm2 after anti-VEGF treatment at six months (P < 0.001). In the subgroup receiving monthly anti-VEGF injections, the global RVBA decreased from 68.7 ± 16.2 to 33.9 ± 13.3 mm2 (P < 0.001). In the subgroup receiving anti-VEGF every three months, the global RVBA decreased from 65.6 ± 15.1 to 50.8 ± 19.3 mm2 (P = 0.004). CONCLUSION: RVBA seems to be a new biomarker to indicate efficiency of retinal vascular changes after anti-VEGF injection. Eyes with proliferative diabetic retinopathy and significant nonperfusion demonstrate reduced RVBA after anti-VEGF treatment.

Factors Affecting Predominantly Peripheral Lesion Identification and Grading.

Purpose: The purpose of this study was to determine factors affecting predominantly peripheral lesion (PPL) grading, such as qualitative versus quantitative assessment, device type, and severity of diabetic retinopathy (DR) in ultrawide field color images (UWF-CIs). Methods: Patients with DR had UWF-CI qualitatively graded for PPL using standardized techniques and had hemorrhages/microaneurysms (H/Mas) individually annotated for quantitative PPL grading on two different ultrawide field devices. Results: Among 791 eyes of 481 patients, 38.2% had mild nonproliferative DR (NPDR), 34.7% had moderate NPDR, and 27.1% had severe NPDR to proliferative DR (PDR). The overall agreement between qualitative and quantitative PPL grading was moderate (ĸ = 0.423, P < 0.001). Agreement rates were fair in eyes with mild NPDR (ĸ = 0.336, P < 0.001) but moderate in eyes with moderate NPDR (ĸ = 0.525, P < 0.001) and severe NPDR-PDR (ĸ = 0.409, P < 0.001). Increasing thresholds for quantitative PPL determination improved agreement rates, with peak agreements at H/Ma count differences of six for mild NPDR, five for moderate NPDR, and nine for severe NPDR-PDR. Based on ultrawide field device type (California = 412 eyes vs. 200Tx = 379 eyes), agreement between qualitative and quantitative PPL grading was moderate for all DR severities in both devices (ĸ = 0.369-0.526, P < 0.001) except for mild NPDR on the 200Tx, which had poor agreement (ĸ = 0.055, P = 0.478). Conclusions: Determination of PPL varies between standard qualitative and quantitative grading and is dependent on NPDR severity, device type, and magnitude of lesion differences used for quantitative assessment. Translational Relevance: Prior UWF studies have not accounted for imaging and grading factors that affect PPL, such factors need to be reviewed when assessing thresholds for DR progression rates.

Automated Microaneurysm Counts on Ultrawide Field Color and Fluorescein Angiography Images.

BACKGROUND: The severity and extent of microaneurysms (MAs) have been used to determine diabetic retinopathy (DR) severity and estimate the risk of DR progression over time. The recent introduction of ultrawide field (UWF) imaging has allowed ophthalmologists to readily image nearly the entire retina. Manual counting of MAs, especially on UWF images, is laborious and time-consuming, limiting its potential use in clinical settings. Automated MA counting techniques are potentially more accurate and reproducible compared to manual methods. METHOD: Review of available literature on current techniques of automated MA counting techniques on both ultrawide field (UWF) color images (CI) and fluorescein angiography (FA) images. RESULTS: Automated MA counting techniques on UWF images are still in the early phases of development with UWF-FA counts being further along. Early studies have demonstrated that these techniques are accurate and reproducible. CONCLUSION: Automated techniques may be an appropriate option for detecting and quantifying MAs on UWF images, especially in eyes with earlier DR severity. Larger studies are needed to appropriately validate these techniques and determine if they add substantially to clinical practice compared to standard DR grading.

Retinal Vascular Caliber Association with Nonperfusion and Diabetic Retinopathy Severity Depends on Vascular Caliber Measurement Location.

PURPOSE: To evaluate the association of retinal nonperfusion and diabetic retinopathy (DR) severity with location of vascular caliber measurement using ultrawide field (UWF) imaging. DESIGN: Retrospective image review. PARTICIPANTS: Adults with diabetes mellitus. METHODS: All images from subjects with same-day UWF fluorescein angiography (FA) and color imaging were evaluated. Predominantly peripheral lesions (PPL) and DR severity were graded from UWF color images. Nonperfusion was quantified using UWF-FA in defined retinal regions [posterior pole (PP), mid-periphery (MP), far-periphery (FP)]. Retinal vessel calibers were measured at an optic disc centered inner and outer zone. MAIN OUTCOME MEASURES: Nonperfusion index (NPI) in the PP, MP and FP. Mean arteriole and venule diameter in the inner and outer zones. RESULTS: Two hundred eighty-five eyes of 193 patients (24.9% mild nonproliferative DR [NPDR], 22.8% moderate NPDR, 37.5% severe NPDR and 14.7% proliferative DR [PDR]) were reviewed. No significant associations between inner zone arteriolar diameter and retinal NPI overall or in any retinal region. In the outer zone, eyes with thinnest arteriolar calibers (quartile 1) were associated with a 1.7- to 2.4-fold nonperfusion increase across all retinal regions compared to the remaining eyes (P = 0.002 [PP] to 0.048 [FP]). In the outer zone, the percentage of eyes in the thinnest quartile of retinal arteriolar diameter increased with worsening DR severity (mild NPDR: 10% vs PDR: 31%, P = 0.007). This association was not observed when measured within the inner zone (P = 0.129). All venular caliber associations were not statistically significant when corrected for potentially confounding factors. Thinner outer zone retinal arteriolar caliber (quartile 1) was more common in eyes with PPL compared to eyes without PPL (34.1% vs 20.8%, P = 0.017) as were thicker outer venular calibers (quartile 4) (33% vs 21.3%, P = 0.036). Presence of PPL was associated with thinner outer zone arteriolar caliber (109.7 ± 26.5µm vs 123.0 ± 29.5µm, P = 0.001). CONCLUSIONS: The association of vascular caliber with nonperfusion and DR severity differs based upon the retinal location at which vascular caliber is measured. Peripheral arterial narrowing is associated with increasing nonperfusion, worsening DR severity and presence of PPL. In contrast, inner zone retinal arteriolar caliber is not associated with these findings.

Fractal analysis of retinal vasculature in normal subjects on ultra-wide field fluorescein angiography.

AIM: To evaluate the fractal feature of the retinal vasculature of normal eyes on a stereographic projected and montaged ultra-wide field (UWF) fluorescein angiography (FA). METHODS: Prospective, observational, cross-sectional study. Totally 59 eyes of 31 normal subjects were imaged using the Optos 200Tx. Images obtained at different gaze angles stereographically projected and montaged. The early-phase UWF FA frames were processed to segment the retinal vasculature and the results were exported as binary masks. The fractal dimension (FD) was calculated using the box-counting method. RESULTS: The global FD for the entire retina was 1.6±0.04, with no difference between males and females (1.59±0.04 vs 1.61±0.04, P=0.084) or between right and left eyes (1.6±0.04 vs 1.6±0.05, P=0.61). FD was non-uniformly distributed among four quadrants (P<0.001) and decreased as the distance from the fovea increased (P<0.001). A negative association was observed between FD and age (R=-0.37, P=0.006), and this relationship was observed in the posterior and mid-peripheral retina (P<0.05) but absent in far-periphery (P>0.05). CONCLUSION: Fractal geometry is non-uniformly distributed across the retina in normal eyes and decreases from the fovea to the far-periphery. Subjects with an older age tend to have a smaller FD, however, the FD in the far-periphery does not appear to be influenced by age.

Disparity of microaneurysm count between ultrawide field colour imaging and ultrawide field fluorescein angiography in eyes with diabetic retinopathy.

AIMS: To compare microaneurysm (MA) counts using ultrawide field colour images (UWF-CI) and ultrawide field fluorescein angiography (UWF-FA). METHODS: Retrospective study including patients with type 1 or 2 diabetes mellitus receiving UWF-FA and UWF-CI within 2 weeks. MAs were manually counted in individual Early Treatment Diabetic Retinopathy Study (ETDRS) and extended UWF zones. Fields with MAs ≥20 determined diabetic retinopathy (DR) severity (0 fields=mild, 1-3=moderate, ≥4=severe). UWF-FA and UWF-CI agreement was determined and UWF-CI DR severity sensitivity analysis adjusting for UWF-FA MA counts performed. RESULTS: In 193 patients (288 eyes), 2.4% had no DR, 29.9% mild non-proliferative DR (NPDR), 32.6% moderate (NPDR), 22.9% severe NPDR and 12.2% proliferative DR. UWF-FA MA counts were 3.5-fold higher (p<0.001) than UWF-CI counts overall, 3.2x-fold higher in ETDRS fields (p<0.001) and 5.3-fold higher in extended ETDRS fields (p<0.001) and higher in type 1 versus type 2 diabetes (p<0.001). In eyes with NPDR on UWF-CI (n=246), UWF-FA images had 1.6x-3.5x more fields with ≥20 MAs (p<0.001). Fair agreement existed between imaging modalities (k=0.221-0.416). In ETDRS fields, DR severity agreement increased from k=0.346 to 0.600 when dividing UWF-FA counts by a factor of 3, followed by rapid decline in agreement thereafter. Total UWF area agreement increased from k=0.317 to 0.565 with an adjustment factor of either 4 or 5. CONCLUSIONS: UWF-FA detects threefold to fivefold more MAs than UWF-CI and identifies 1.6-3.5-fold more fields affecting DR severity. Differences exist at all DR severity levels, thus limiting direct comparison between the modalities. However, correcting UWF-FA MA counts substantially improves DR severity agreement between the modalities.

Association between hypertension and retinal vascular features in ultra-widefield fundus imaging.

Objective: Changes to the retinal vasculature are known to be associated with hypertension independently of traditional risk factors. We investigated whether measurements of retinal vascular calibre from ultra-widefield fundus imaging were associated with hypertensive status. Methods: We retrospectively collected and semiautomatically measured ultra-widefield retinal fundus images from a subset of participants enrolled in an ongoing population study of ageing, categorised as normotensive or hypertensive according to thresholds on systolic/diastolic blood pressure (140/90 mm Hg) measured in a clinical setting. Vascular calibre in the peripheral retina was measured to calculate the nasal-annular arteriole:venule ratio (NA-AVR), a novel combined parameter. Results: Left and right eyes were analysed from 440 participants (aged 50-59 years, mean age of 54.6±2.9 years, 247, 56.1% women), including 151 (34.3%) categorised as hypertensive. Arterioles were thinner and the NA-AVR was smaller in people with hypertension. The area under the receiver operating characteristic curve of NA-AVR for hypertensive status was 0.73 (95% CI 0.68 to 0.78) using measurements from left eyes, while for right eyes, it was 0.64 (95% CI 0.59 to 0.70), representing evidence of a statistically significant difference between the eyes (p=0.020). Conclusions: Semiautomated measurements of NA-AVR in ultra-widefield fundus imaging were associated with hypertension. With further development, this may help screen people attending routine eye health check-ups for high blood pressure. These individuals may then follow a care pathway for suspected hypertension. Our results showed differences between left and right eyes, highlighting the importance of investigating both eyes of a patient.

Relationship Between Retinal Fractal Dimension and Nonperfusion in Diabetic Retinopathy on Ultrawide-Field Fluorescein Angiography.

PURPOSE: To correlate fractal dimension (FD) of the retinal vasculature with the extent of retinal nonperfusion area in diabetic retinopathy (DR) on ultrawide-field fluorescein angiography (FA). DESIGN: Cross-sectional study. METHODS: Baseline Optos 200Tx ultrawide-field FA images of 80 eyes with DR from the DAVE (NCT01552408) and RECOVERY (NCT02863354) studies were stereographically projected at the Doheny Image Reading Center. The retinal vasculature was extracted from an early-phase FA frame by exploiting the elongated nature of the vessels and then skeletonized for calculation of FD using a box-counting method. The nonperfusion area was delineated by 2 independent, reading center-certified graders who were masked to the study groups and who were using a standardized protocol and then computed in millimeters squared. RESULTS: While no difference in FD was observed for the entire retina in DR compared with normal control subjects, a significantly smaller FD was found in the far-periphery of the DR eyes (P < .001). FD for the entire retina was negatively associated with global nonperfusion area (R = -0.44; P < .001), and this relationship was also present within the 3 concentric retinal zones (posterior: R = -0.31, P = .016; midperiphery: R = -0.35, P = .007; and far periphery: R = -0.31, P = .015). CONCLUSIONS: Peripheral FD on ultrawide-field FA is reduced in DR eyes compared with normal eyes and is correlated with severity of retinal nonperfusion. FD can be calculated automatically without the need for correction of peripheral distortion, and therefore it may prove to be a useful surrogate biomarker when precise quantification of nonperfusion is not feasible.

Distribution of Nonperfusion and Neovascularization on Ultrawide-Field Fluorescein Angiography in Proliferative Diabetic Retinopathy (RECOVERY Study): Report 1.

PURPOSE: To explore the distribution of nonperfusion area (NPA) on ultrawide-field fluorescein angiography (UWF FA) in proliferative diabetic retinopathy (PDR) and its relationship with the presence of neovascularization of the optic disc (NVD) and distribution of neovascularization elsewhere (NVE). DESIGN: Prospective, observational case series. METHODS: Baseline Optos 200Tx UWF FA images of 38 eyes with treatment-naïve early-stage PDR from the RECOVERY (NCT02863354) study were stereographically projected at the Doheny Image Reading Center. Two independent/masked certified graders manually delineated the NPA and the total visible retinal area (TRA). NPA and TRA were then computed in square millimeters using the manufacturer software. Ischemic index (ISI) was calculated by dividing NPA by TRA. NPA and ISI were correlated with the presence and distribution of neovascularization in the corresponding zones. RESULTS: Eyes with NVD appeared to have more severe global NPA than those without (P = .026). Although the ISI appeared to increase with increasing distance from the foveal center (P < .001), NVE was more likely to be located in the posterior pole than the midperiphery or far-periphery (P < .001). Presence of NVE in the posterior polar retina appeared to demonstrate more severe ischemia in the posterior pole and midperiphery than those without (P < .05), but interestingly, was not correlated with the severity of overall global ischemia or of ischemia in the far-periphery alone (P > .05). CONCLUSIONS: Whereas the presence of NVD was associated with the severity of global ischemia, the distribution of NVE did not appear to be influenced by the distribution of ischemia.

Automated detection of age-related macular degeneration in color fundus photography: a systematic review.

The rising prevalence of age-related eye diseases, particularly age-related macular degeneration, places an ever-increasing burden on health care providers. As new treatments emerge, it is necessary to develop methods for reliably assessing patients' disease status and stratifying risk of progression. The presence of drusen in the retina represents a key early feature in which size, number, and morphology are thought to correlate significantly with the risk of progression to sight-threatening age-related macular degeneration. Manual labeling of drusen on color fundus photographs by a human is labor intensive and is where automatic computerized detection would appreciably aid patient care. We review and evaluate current artificial intelligence methods and developments for the automated detection of drusen in the context of age-related macular degeneration.

Effect of phase-plate adjustment on retinal image sharpness and visible retinal area on ultrawide field imaging.

BACKGROUND: To evaluate changes in image sharpness across ultrawide field (UWF) images and the effect of phase-plate adjustment on image contrast and extent of visible retinal area (VRA). METHODS: This was a single site evaluation of 200° UWF images acquired with phase-plate adjustment (California, Optos, plc) and without (200TX, Optos, plc). Images were acquired using standardized protocol. VRA was manually outlined on each image and quantified using customized software. Mean image sharpness was evaluated using an automated method within the full VRA of each image and within the peripheral region of the VRA. The VRA and image sharpness were evaluated and compared between the two devices. RESULTS: Twenty eyes of 10 healthy volunteers were evaluated. Devices with and without phase-plate adjustment produced a similar extent of VRA. Eye steering increased VRA in devices with and without phase-plate adjustment by 39.3% and 34.3%, respectively. Regardless of gaze direction, mean sharpness of the full VRA was reduced in peripheral area with or without phase-plate adjustment. Compared to images without phase-plate adjustment, use of phase-plate adjustment reduced the loss of peripheral image sharpness in all fields (-4.2 to -26.0%; p < 0.001 all fields). The sharpness of the peripheral area for on-axis images was 61.5% higher with phase-plate adjustment. CONCLUSIONS: The use of phase-plate adjustment does not alter the extent of VRA. However, for on-axis images the loss of sharpness in the periphery is 4.5-fold less with phase-plate adjustment, potentially reducing the need to steer images and improving lesion detection in these areas.

Comparison of Subjective Assessment and Precise Quantitative Assessment of Lesion Distribution in Diabetic Retinopathy.

Importance: Predominantly peripheral disease in eyes with nonproliferative diabetic retinopathy (DR) is suggested as a potential strong risk factor for progression to proliferative disease. However, the reliability and optimal method for the assessment of lesion distribution are still uncertain. Objective: To compare agreement between subjective assessment and precise quantification of lesion burden in ultrawidefield (UWF) images of eyes with DR. Design, Setting, and Participants: This multisite cross-sectional study examines UWF pseudocolor images acquired from DR screening clinic patients from December 20, 2014, through August 1, 2014. Of 104 cases, 161 eyes with DR were included. Data analysis was conducted from June 1, 2016, through December 1, 2016 at the Doheny Image Reading Center. Main Outcomes and Measures: Distribution of DR lesions in eyes was assessed subjectively and quantitatively, and eyes were classified as having predominantly central lesions (PCLs) or predominantly peripheral lesions (PPLs). The frequency and surface area (SA) of each lesion type were quantified. Intergrader and subjective vs quantitative classification were compared for level of agreement. Several methods of determining PPL distribution were also compared. Results: On subjective frequency-based evaluation by graders, 133 eyes were classified as having PCL, and 28 eyes as having PPL. On exact quantification of lesion SA, 121 eyes were classified as PCL, and 40 eyes as having PPL. On SA-based quantification, 134 eyes were classified as having PCL, and 27 eyes as having PPL. There was a significant difference between qualitative and quantitative classification of DR lesion distribution for both frequency-based (mean difference [SD]: PCL, 6 [2]; PPL, 13 [6]; P < .001) and SA-based (mean difference [SD]: PCL, 6 [1]; PPL, 20 [7]; P < .001) methods. Both intergrader reproducibility and subjective vs quantitative agreement were higher with frequency-based classification. Conclusions and Relevance: Subjective assessment of PPL DR lesions on UWF images differed in some cases from precise quantitative assessments, particularly when considering the area of lesions. These findings highlight the benefit of objective quantitative approaches to DR assessment, which may facilitate the development of a more precise DR scoring system.

A Novel Adaptive Deformable Model for Automated Optic Disc and Cup Segmentation to Aid Glaucoma Diagnosis.

This paper proposes a novel Adaptive Region-based Edge Smoothing Model (ARESM) for automatic boundary detection of optic disc and cup to aid automatic glaucoma diagnosis. The novelty of our approach consists of two aspects: 1) automatic detection of initial optimum object boundary based on a Region Classification Model (RCM) in a pixel-level multidimensional feature space; 2) an Adaptive Edge Smoothing Update model (AESU) of contour points (e.g. misclassified or irregular points) based on iterative force field calculations with contours obtained from the RCM by minimising energy function (an approach that does not require predefined geometric templates to guide auto-segmentation). Such an approach provides robustness in capturing a range of variations and shapes. We have conducted a comprehensive comparison between our approach and the state-of-the-art existing deformable models and validated it with publicly available datasets. The experimental evaluation shows that the proposed approach significantly outperforms existing methods. The generality of the proposed approach will enable segmentation and detection of other object boundaries and provide added value in the field of medical image processing and analysis.

Hemorrhage and/or Microaneurysm Severity and Count in Ultrawide Field Images and Early Treatment Diabetic Retinopathy Study Photography.

OBJECTIVE: To evaluate detection of hemorrhage and/or microaneurysm (H/Ma) using ultrawide field (UWF) retinal imaging as compared with standard Early Treatment Diabetic Retinopathy Study (ETDRS) 7-field photographs (ETDRS photos). DESIGN: Single-site comparative study of UWF images and ETDRS photos. PARTICIPANTS: One hundred twenty-six eyes of 69 patients with no diabetic retinopathy (DR) or mild or moderate nonproliferative DR (NPDR). METHODS: Stereoscopic 200° UWF images and stereoscopic 35mm 30° 7-field color photographs were acquired on the same visit. Images were graded for severity and distribution of H/Ma. H/Mas were counted in ETDRS fields 2 to 7 in both ETDRS photos and UWF images. H/Mas in the UWF peripheral fields were also counted. MAIN OUTCOME MEASURES: Kappa (κ) and weighted κ statistics for agreement. Number of H/Ma within and outside ETDRS fields identified in UWF images and ETDRS photos. RESULTS: Distribution of DR severity by ETDRS photos was 24 (19.0%) no DR, 48 (38.1%) mild NPDR, and 54 (42.9%) moderate NPDR. A total of 748 of 756 fields (98.9%) were gradable for H/Mas on ETDRS photos and UWF images. Simple κ/weighted κ statistics for severity of H/Ma: all fields 0.61/0.69, field 2 0.70/0.77, field 3 0.62/0.73, field 4 0.50/0.62, field 5 0.54/0.65, field 6 0.64/0.70, and field 7 0.58/0.63 with overall exact agreement in 81.3% and within 1 step in 97.9% of fields. A greater proportion of fields was graded a more severe H/Ma level in UWF images than in the corresponding ETDRS photos (UWF: 12.7% vs. ETDRS: 6.5%). Evaluating comparable areas in UWF images and ETDRS photos (fields 2-7), a mean of 42.8 H/Mas were identified using ETDRS photos and 48.8 in UWF images (P = 0.10). An additional mean of 21.3 H/Mas (49.8% increase, P < 0.0001) were identified in the peripheral fields of the UWF images. CONCLUSIONS: There is good to excellent agreement between UWF images and ETDRS photos in determining H/Ma severity, with excellent correlation of H/Ma counts within ETDRS photo fields. UWF peripheral fields identified 49.8% more H/Ma, suggesting a more severe H/Ma in 12.7% of eyes. Given the additional lesions detected in peripheral fields and the known risks associated with H/Ma and peripheral lesions, quantification of H/Ma using UWF images may provide a more accurate representation of DR disease activity and potential greater accuracy in predicting DR progression.