Automated Detection of Glaucoma With Interpretable Machine Learning Using Clinical Data and Multimodal Retinal Images

PURPOSE

To develop a multimodal model to automate glaucoma detection DESIGN: Development of a machine-learning glaucoma detection model METHODS: We selected a study cohort from the UK Biobank data set with 1193 eyes of 863 healthy subjects and 1283 eyes of 771 subjects with glaucoma. We trained a multimodal model that combines multiple deep neural nets, trained on macular optical coherence tomography volumes and color fundus photographs, with demographic and clinical data. We performed an interpretability analysis to identify features the model relied on to detect glaucoma. We determined the importance of different features in detecting glaucoma using interpretable machine learning methods. We also evaluated the model on subjects who did not have a diagnosis of glaucoma on the day of imaging but were later diagnosed (progress-to-glaucoma [PTG]).

RESULTS

Results show that a multimodal model that combines imaging with demographic and clinical features is highly accurate (area under the curve 0.97). Interpretation of this model highlights biological features known to be related to the disease, such as age, intraocular pressure, and optic disc morphology. Our model also points to previously unknown or disputed features, such as pulmonary function and retinal outer layers. Accurate prediction in PTG highlights variables that change with progression to glaucoma-age and pulmonary function.

CONCLUSIONS

The accuracy of our model suggests distinct sources of information in each imaging modality and in the different clinical and demographic variables. Interpretable machine learning methods elucidate subject-level prediction and help uncover the factors that lead to accurate predictions, pointing to potential disease mechanisms or variables related to the disease.

Reduced Pulsatile Trabecular Meshwork Motion in Eyes With Primary Open Angle Glaucoma Using Phase-Sensitive Optical Coherence Tomography

Purpose

The purpose of this study was to investigate the difference in pulsatile trabecular meshwork (TM) motion between normal and eyes with POAG using phase-sensitive optical coherence tomography (PhS-OCT).

Methods

In this cross-sectional study, eight healthy subjects (16 eyes) and nine patients with POAG (18 eyes) were enrolled. A laboratory-based prototype PhS-OCT system was used to measure pulsatile TM motion. PhS-OCT images were analyzed to obtain parameters of pulsatile TM motion (i.e. maximum velocity [MV] and cumulative displacement [CDisp]). Outflow facility and ocular pulse amplitude were measured using pneumotonography. Detection sensitivity was compared among various parameters by calculating the area under the receiver operating characteristic curves (AUCs).

Results

A pulsatile TM motion waveform synchronous with digital pulse was observed using PhS-OCT in both healthy and POAG eyes. The mean MV in eyes with glaucoma was significantly lower than healthy eyes (P < 0.001). The mean CDisp in POAG eyes was also significantly lower than healthy eyes (P < 0.001). CDisp showed a significant correlation (r = 0.46; P = 0.0088) with ocular pulse amplitude in the study. Compared with the outflow facility, both the MV and CDisp were found to have a better discrimination of glaucoma (P < 0.001 and P = 0.0074, respectively).

Conclusions

Pulsatile TM motion was reduced in patients with POAG compared to healthy subjects. The underlying mechanism may be due to the altered tissue stiffness or other biomechanical properties of the TM in POAG eyes. Our evidence suggests that the measurement of pulsatile TM motion with PhS-OCT may help in characterizing outflow pathway abnormalities.

Laser Peripheral Iridotomy Curriculum: Lecture and Simulation Practical

INTRODUCTION

Approximately 20 million people worldwide are affected by primary angle closure glaucoma, which is often treated with a laser peripheral iridotomy (LPI). In the United States, at least 60,000 to 80,000 LPIs are performed annually. However, complications can arise from improperly performed LPIs. While the Accreditation Council for Graduate Medical Education requires that all ophthalmology residents perform at least four primary LPIs prior to graduating, formal training is often lacking. In an effort to standardize LPI teaching, an LPI lecture curriculum and skills practice session were introduced.

METHODS

A lecture and wet-lab curriculum was developed at the University of Washington to formally teach first-year ophthalmology residents the indications and techniques for LPI. Pre- and postcurriculum knowledge was tested, and LPI performance was assessed by comparing pre- and postcurriculum total number of shots and time needed to successfully complete an LPI on a commercially available model eye.

RESULTS

The course was highly rated by 10 residents (all PGY 2), with an increase in pre- versus posttest scores, an improvement in LPI performance metrics, and an increase in pre- versus postcurriculum scores for the three survey questions regarding curriculum objectives.

DISCUSSION

This course improved learner knowledge and confidence in performing LPI. Test scores improved following the course, as did self-assessed confidence levels of the residents. Residents made a number of positive comments about the course. We plan to continue holding this training session every year at our institution.

Using Deep Learning to Automate Goldmann Applanation Tonometry Readings

PURPOSE

To develop an objective and automated method for measuring intraocular pressure using deep learning and fixed-force Goldmann applanation tonometry (GAT) techniques.

DESIGN

Prospective cross-sectional study.

PARTICIPANTS

Patients from an academic glaucoma practice.

METHODS

Intraocular pressure was estimated by analyzing videos recorded using a standard slit-lamp microscope and fixed-force GAT. Video frames were labeled to identify the outline of the reference tonometer and the applanation mires. A deep learning model was trained to localize and segment the tonometer and mires. Intraocular pressure values were calculated from the deep learning-predicted tonometer and mire diameters using the Imbert-Fick formula. A separate test set was collected prospectively in which standard and automated GAT measurements were collected in random order by 2 independent masked observers to assess the deep learning model as well as interobserver variability.

MAIN OUTCOME MEASURES

Intraocular pressure measurements between standard and automated methods were compared.

RESULTS

Two hundred sixty-three eyes of 135 patients were included in the training and validation videos. For the test set, 50 eyes from 25 participants were included. Each eye was measured by 2 observers, resulting in 100 videos. Within the test set, the mean difference between automated and standard GAT results was -0.9 mmHg (95% limits of agreement [LoA], -5.4 to 3.6 mmHg). Mean difference between the 2 observers using standard GAT was 0.09 mmHg (LoA,-3.8 to 4.0 mmHg). Mean difference between the 2 observers using automated GAT videos was -0.3 mmHg (LoA, -4.1 to 3.5 mmHg). The coefficients of repeatability for automated and standard GAT were 3.8 and 3.9 mmHg, respectively. The bias for even-numbered measurements was reduced when using automated GAT.

CONCLUSIONS

Preliminary measurements using deep learning to automate GAT demonstrate results comparable with those of standard GAT. Automated GAT has the potential to improve on our current GAT measurement standards significantly by reducing bias and improving repeatability. In addition, ocular pulse amplitudes could be observed using this technique.

Development and validation of a machine learning, smartphone-based tonometer

BACKGROUND/AIMS

To compare intraocular pressure (IOP) measurements using a prototype smartphone tonometer with other tonometers used in clinical practice.

METHODS

Patients from an academic glaucoma practice were recruited. The smartphone tonometer uses fixed force applanation and in conjunction with a machine-learning computer algorithm is able to calculate the IOP. IOP was also measured using Goldmann applanation tonometry (GAT) in all subjects. A subset of patients were also measured using ICare, pneumotonometry (upright and supine positions) and Tono-Pen (upright and supine positions) and the results were compared.

RESULTS

92 eyes of 81 subjects were successfully measured. The mean difference (in mm Hg) for IOP measurements of the smartphone tonometer versus other devices was +0.24 mm Hg for GAT, -1.39 mm Hg for ICare, -3.71 mm Hg for pneumotonometry and -1.30 mm Hg for Tono-Pen. The 95% limits of agreement for the smartphone tonometer versus other devices was -4.35 to 4.83 mm Hg for GAT, -6.48 to 3.70 mm Hg for ICare, -7.66 to -0.15 mm Hg for pneumotonometry and -5.72 to 3.12 mm Hg for Tono-Pen. Overall, the smartphone tonometer results correlated best with GAT (R²=0.67, p<0.001). Of the 92 videos, 90 (97.8%) were within ±5 mm Hg of GAT and 58 (63.0%) were within ±2 mm Hg of GAT.

CONCLUSIONS

Preliminary IOP measurements using a prototype smartphone-based tonometer was grossly equivalent to the reference standard.

Effect of prior glaucoma surgery on intraocular pressure immediately after anti-vascular endothelial growth factor injection

BACKGROUND

To characterize how prior incisional glaucoma surgery affects the intraocular pressure (IOP) elevation immediately following intravitreal anti-VEGF injections (IVI).

METHODS

Single institution, experimental study. Patients with a history of incisional glaucoma surgery who were receiving anti-VEGF injections were recruited as well as control eyes. Pre- and post-injection IOP measurements were compared as well as time to recovery to within 5 and 10 mmHg of baseline IOP.

RESULTS

Ten eyes with a history of glaucoma surgery and 29 control eyes receiving anti-VEGF injections were included. The most common indication for intravitreal anti-VEGF injection was proliferative diabetic retinopathy in both surgical and control eyes (50% vs 45%, p = 1.00). Post-injection IOP was significantly decreased compared to baseline IOP after anti-VEGF injection in surgical versus control eyes (26.5 ± 8.9 mmHg vs 44.2 ± 8.5 mmHg, respectively, p < 0.001). The mean change in IOP following intravitreal anti-VEGF injection was lower in surgical eyes (10.7 ± 6.6 mmHg vs 28.6 ± 8.3 mmHg, p < 0.001). The mean time for the IOP to return to within 10 mmHg of pre-injection IOP was less in surgical eyes (5.2 ± 4.1 min vs 13.3 ± 7.6 min, p = 0.002).

CONCLUSIONS

Eyes with prior incisional glaucoma surgery demonstrated a significantly lower post-injection IOP elevation and a faster recovery to within 10 mmHg of their pre-injection IOP. Incisional glaucoma surgery may be considered for patients where the attenuation of post-injection IOP elevation is needed and other less invasive measures have failed.

Forecasting future Humphrey Visual Fields using deep learning

Purpose

To determine if deep learning networks could be trained to forecast future 24–2 Humphrey Visual Fields (HVFs).

Methods

All data points from consecutive 24–2 HVFs from 1998 to 2018 were extracted from a university database. Ten-fold cross validation with a held out test set was used to develop the three main phases of model development: model architecture selection, dataset combination selection, and time-interval model training with transfer learning, to train a deep learning artificial neural network capable of generating a point-wise visual field prediction. The point-wise mean absolute error (PMAE) and difference in Mean Deviation (MD) between predicted and actual future HVF were calculated.

Results

More than 1.7 million perimetry points were extracted to the hundredth decibel from 32,443 24–2 HVFs. The best performing model with 20 million trainable parameters, CascadeNet-5, was selected. The overall point-wise PMAE for the test set was 2.47 dB (95% CI: 2.45 dB to 2.48 dB), and deep learning showed a statistically significant improvement over linear models. The 100 fully trained models successfully predicted future HVFs in glaucomatous eyes up to 5.5 years in the future with a correlation of 0.92 between the MD of predicted and actual future HVF and an average difference of 0.41 dB.

Conclusions

Using unfiltered real-world datasets, deep learning networks show the ability to not only learn spatio-temporal HVF changes but also to generate predictions for future HVFs up to 5.5 years, given only a single HVF.

Resident-performed laser peripheral iridotomy in primary angle closure, primary angle closure suspects, and primary angle closure glaucoma

Purpose

To investigate the power use and complication frequency of resident-performed laser peripheral iridotomy (LPI).

Methods

A retrospective analysis of 196 eyes from 103 patients who underwent neodymium: yttrium-aluminum-garnet laser iridotomy performed by resident physicians from January 1, 2010 through April 30, 2015 at a university-based county hospital was done. All patients were treated for primary angle closure, primary angle closure suspects, and primary angle closure glaucoma. Data were collected on pre- and post-laser intraocular pressure (IOP), ethnicity, laser parameters and complications. Mean power use and frequency of complications were evaluated. Complications included elevated post-laser IOP at 30–45 minutes (≥8 mmHg), hyphema, aborted procedures, and lasering non-iris structures. The number of repeated LPI procedures, was also recorded.

Results

Mean total power used for all residents was 78.2±68.7 mJ per eye. Power use by first-year trainees was significantly higher than second- and third-year trainees (103.5±75.5 mJ versus 73.7±73.8 mJ and 67.2±56.4 mJ, respectively, p=0.011). Complications included hyphema or microhyphema in 17.9% (35/196), IOP spikes in 5.1% (10/196), aborted procedures in 1.1% (3/196) and lasering non-iris structures in 0.5% (1/196). LPI was repeated in 22.4% of cases (44/196) with higher incidence of repeat LPI among non-Caucasian compared to the Caucasian subjects (p=0.02). Complication rates did not differ with increased training (p=0.16).

Conclusion

Total power used for LPI decreased with increased resident training, while the complication rate did not differ significantly among resident classes. Complication rates were comparable to rates reported in the literature for attending-performed LPIs.

Peripapillary Retinal Nerve Fiber Layer Vascular Microcirculation in Eyes With Glaucoma and Single-Hemifield Visual Field Loss

Importance

Understanding the differences in vascular microcirculation of the peripapillary retinal nerve fiber layer (RNFL) between the hemispheres in eyes with glaucoma with single-hemifield visual field (VF) defects may provide insight into the pathophysiology of glaucoma.

Objective

To investigate the changes in the microcirculation of the peripapillary RNFL of eyes with glaucoma by using optical microangiography.

Design, Setting, and Participants

Eyes with glaucoma and single-hemifield VF defect and normal eyes underwent scanning using an optical microangiography system covering a 6.7 × 6.7-mm2 area centered at the optic nerve head. The RNFL microcirculation was measured within an annulus region centered at the optic nerve head divided into superior and inferior hemispheres. Blood flux index (the mean flow signal intensity in the vessels) and vessel area density (the percentage of the detected vessels in the annulus) were measured.

Main Outcomes and Measures

Differences in microcirculation between the hemispheres in eyes with glaucoma and normal eyes and correlations among blood flow metrics, VF thresholds, and clinical optical coherence tomography structural measurements were assessed.

Results

Twenty-one eyes from 21 patients with glaucoma (7 men and 14 women; mean [SD] age, 63.7 [9.9] years) and 20 eyes from 20 healthy control individuals (9 men and 11 women; mean [SD] age, 68.3 [10.7] years) were studied. In eyes with glaucoma, the abnormal hemisphere showed a thinner RNFL (mean [SE] difference, 23.5 [4.5] μm; 95% CI, 15.1-32.0 µm; P < .001), lower RNFL blood flux index (mean [SE] difference, 0.04 [0.01]; 95% CI, 0.02-0.05; P < .001), and less vessel area density (mean [SE] difference, 0.08% [0.02%]; 95% CI, 0.05%-0.10%; P < .001) than did the normal hemisphere. Compared with normal eyes, reduced RNFL microcirculation was found in the normal hemisphere of eyes with glaucoma, measured by mean [SE] differences in blood flux index (0.06 [0.01]; 95% CI, 0.04-0.09; P < .001) and vessel area density (0.04% [0.02%]; 95% CI, 0.02%-0.08%; P = .003) but not in RNFL thickness (3.4 [4.7] μm; 95% CI, -6.2 to 12.9 µm; P = .48). Strong correlations were found between the blood flux index and VF mean deviation (Spearman ρ = 0.44; P = .045) and RNFL thickness (Spearman ρ = 0.65; P = .001) in the normal hemisphere of the eye with glaucoma.

Conclusions and Relevance

Reduced RNFL microcirculation was detected in the normal hemisphere of eyes with glaucoma, with strong correspondence with VF loss and RNFL thinning. Although the results suggest that vascular dysfunction precedes structural changes seen in glaucoma, longitudinal studies would be needed to confirm this finding.

VEGF as a Paracrine Regulator of Conventional Outflow Facility

Purpose

Vascular endothelial growth factor (VEGF) regulates microvascular endothelial permeability, and the permeability of Schlemm's canal (SC) endothelium influences conventional aqueous humor outflow. We hypothesize that VEGF signaling regulates outflow facility.

Methods

We measured outflow facility (C) in enucleated mouse eyes perfused with VEGF-A₁₆₄a, VEGF-A₁₆₅b, VEGF-D, or inhibitors to VEGF receptor 2 (VEGFR-2). We monitored VEGF-A secretion from human trabecular meshwork (TM) cells by ELISA after 24 hours of static culture or cyclic stretch. We used immunofluorescence microscopy to localize VEGF-A protein within the TM of mice.

Results

VEGF-A₁₆₄a increased C in enucleated mouse eyes. Cyclic stretch increased VEGF-A secretion by human TM cells, which corresponded to VEGF-A localization in the TM of mice. Blockade of VEGFR-2 decreased C, using either of the inhibitors SU5416 or Ki8751 or the inactive splice variant VEGF-A₁₆₅b. VEGF-D increased C, which could be blocked by Ki8751.

Conclusions

VEGF is a paracrine regulator of conventional outflow facility that is secreted by TM cells in response to mechanical stress. VEGF affects facility via VEGFR-2 likely at the level of SC endothelium. Disruption of VEGF signaling in the TM may explain why anti-VEGF therapy is associated with decreased outflow facility and sustained ocular hypertension.

Intravitreal Anti-VEGF Injections Reduce Aqueous Outflow Facility in Patients With Neovascular Age-Related Macular Degeneration

Purpose

We assess the effect of intravitreal anti-VEGF injections on tonographic outflow facility.

Methods

Patients with age-related macular degeneration who had received unilateral intravitreal anti-VEGF injections were recruited into two groups, those with ≤10 and those with ≥20 total anti-VEGF injections. Intraocular pressure and tonographic outflow facility of injected and uninjected fellow eyes were measured and compared between groups. Risk factors for development of reduced outflow facility also were assessed.

Results

Outflow facility was 12% lower in the injected eyes of patients who received ≥20 anti-VEGF injections, compared to contralateral uninjected eyes (P = 0.02). In contrast, there was no facility reduction for patients with ≤10 anti-VEGF injections (P = 0.4). In patients with ocular hypertension in the uninjected eye (IOP > 21 mm Hg, n = 5), the outflow facility of injected eyes was on average 46% lower (P = 0.01) than in the uninjected fellow eyes. This was significantly greater than the difference observed in patients with IOP ≤ 21 mm Hg in the uninjected eye (P = 2 × 10⁻⁴). In patients with ocular hypertension in the injected eye (n = 6) the differences in facility and IOP between contralateral eyes were significantly greater than in patients with IOP ≤ 21 mm Hg in the injected eye (P = 2 × 10⁻⁴ and P = 7 × 10⁻⁴, respectively).

Conclusions

Chronic anti-VEGF injections significantly reduce outflow facility in patients with AMD. The greatest facility reduction is observed in patients with baseline ocular hypertension. Ophthalmologists who administer anti-VEGF injections should be aware of these findings and monitor patients closely for changes in IOP or evidence of glaucoma, especially in those with pre-existing ocular hypertension.

Peripapillary Retinal Nerve Fiber Layer Vascular Microcirculation in Glaucoma Using Optical Coherence Tomography-Based Microangiography

Purpose

To investigate the vascular microcirculation changes in the retinal nerve fiber layer (RNFL) in normal, glaucoma suspect, and open-angle glaucoma (OAG) groups using optical coherence tomography–based microangiography (OMAG).

Methods

One eye from each subject was scanned with a Cirrus HD-OCT 5000–based OMAG prototype system montage scanning protocol centered at the optic nerve head (ONH). Blood flow signals were extracted using OMAG algorithm. Retinal nerve fiber layer vascular microcirculation was measured by calculating the blood flux index and vessel area density within a 1.2-mm width annulus centered at the ONH with exclusion of big retinal vessels. One-way ANOVA were performed to analyze the RNFL microcirculation among groups. Linear-regression models were constructed to analyze the correlation between RNFL microcirculation and clinical parameters. Discrimination capabilities of the flow metrics were assessed with the area under the receiver operating characteristic curve (AROC).

Results

Twenty normal, 26 glaucoma suspect, and 42 OAG subjects were enrolled. Eyes from OAG subjects and glaucoma suspects showed significantly lower blood flux index compared with normal eyes (P ≤ 0.0015). Retinal nerve fiber layer blood flow metrics showed significant correlations with visual field indices and structural changes in glaucomatous eyes (P ≤ 0.0123). Similar discrimination capability of blood flux index compared with RNFL thickness was found in both disease groups.

Conclusions

Peripapillary RNFL vascular microcirculation measured as blood flux index by OMAG showed significant differences among OAG, glaucoma suspect, and normal controls and was significantly correlated with functional and structural defects. Retinal nerve fiber layer microcirculation measurement using OMAG may help physicians monitor glaucoma.

A smartphone-based system for assessing intraocular pressure

The measurement of intraocular pressure (IOP) is an important vital sign for the eye, particularly for the diagnosis of glaucoma. Procedures for measuring IOP have been used by eye care professionals for over 100 years, but those without access to such professionals often go undiagnosed. We present a smartphone-based system that can be operated by minimally trained users to measure IOP. The system emulates fixed-force tonometry using a low-cost mechanical attachment to the smartphone. Video is captured through the attachment and then processed in real-time to provide an absolute estimate of the patient's intraocular pressure. Our preliminary assessment with two ex vivo porcine eyes demonstrates that the system follows a baseline physical model with correlations of 0.89 and 0.88.

Cytokine biomarkers in tear film for primary open-angle glaucoma

Purpose

To determine the utility of tear film cytokines as biomarkers for early primary open-angle glaucoma (POAG).

Methods

Patients without POAG and eye drop-naïve patients with newly diagnosed POAG were recruited from an academic hospital-based glaucoma practice. Tear films of recruited patients were obtained and analyzed using a multiplex, high-sensitivity electrochemiluminescent enzyme-linked immunosorbent assay for proinflammatory cytokines (IFNγ, IL-10, IL-12p70, IL-13, IL-1β, IL-2, IL-4, IL-6, IL-8, and TNFα).

Results

Mean concentrations of tear film cytokines were lower in the glaucoma group for 8 of 10 cytokines tested. IL-12p70 (3.94±2.19 pg/mL in control vs 2.31±1.156 pg/mL in POAG; P=0.035) was significantly lower in the tear film of patients with newly diagnosed POAG.

Conclusion

Proinflammatory cytokines were lower in eye drop-naïve newly diagnosed glaucoma patients. Tear film cytokine profiles may be used as biomarkers of early POAG.

Intravitreal Injection Technique

INTRODUCTION

Intravitreal injection of therapeutic medications has become one of the most commonly performed procedures in ophthalmology. Over the past decade, a number of guidelines have been published that recommend proper techniques to increase the safety of intravitreal injections. In this course, we teach and practice the skills necessary to perform safe intravitreal injections.

METHODS

The course was developed for first-year ophthalmology residents and was designed to be delivered in one 2-hour session. Associated materials include guidelines for faculty facilitators and residents to prepare them for safe intravitreal injection of therapeutic medications. Also included are a slide presentation, wet-lab practice guidelines, and an evaluation form.

RESULTS

Learner response has been universally very positive. Residents (N = 15) have reported that structured lectures with practice in the wet lab improved their skill for intravitreal injection of therapeutic medication.

DISCUSSION

Overall, we found that our curriculum improved learner knowledge about safe intravitreal injections and learners' self-assessed confidence in the three objectives of the curriculum.

An upright eyedrop bottle: accuracy, usage of excess drops, and contamination compared to a conventional bottle

Purpose

This study tested the feasibility of using an upright eyedrop bottle (UEB), a device designed to assist patients with eyedrop placement without reclining their head.

Patients and methods

Experienced eyedrop users were enrolled who answered “yes” to the question, “Do you ever have trouble getting your eyedrops in?” After being shown a multimedia presentation and answering a questionnaire regarding eyedrop usage, participants were observed instilling eyedrops. Participants were instructed to instill a single eyedrop in each eye with both a standard bottle and the UEB. They repeated this process three times. With each trial, the amount of time taken to instill drops was recorded, as well as whether a drop landed in the eye (accuracy), if excess drops were used, and if the bottle tip was contaminated.

Results

Forty participants were enrolled, with an average age of 72.4±8.9 years; the majority were females (24 females). Thirty-four participants had been using eyedrops for at least 1 year. The time required to instill eyedrops was significantly less with the UEB in the second and third trials. There was no difference in accuracy between the conventional bottle and the UEB in the left or right eye in any trials. Significantly more participants used excess number of drops while using the conventional bottle in both the left and right eyes in all three trials. The bottle tip was never contaminated with the UEB. Depending on the trial and the eye, the conventional bottle was contaminated by between 42% and 53% of participants.

Conclusion

The UEB has the potential to assist patients with eyedrop placement. Although there was no difference in accuracy between the UEB and the conventional bottle, the UEB was associated with less use of excess drops and less contamination of the bottle tip, compared to the conventional bottle.

Repeatability and reproducibility of optic nerve head perfusion measurements using optical coherence tomography angiography

Optical coherence tomography angiography (OCTA) has increasingly become a clinically useful technique in ophthalmic imaging. We evaluate the repeatability and reproducibility of blood perfusion in the optic nerve head (ONH) measured using optical microangiography (OMAG)-based OCTA. Ten eyes from 10 healthy volunteers are recruited and scanned three times with a 68-kHz Cirrus HD-OCT 5000-based OMAG prototype system (Carl Zeiss Meditec Inc., Dublin, California) centered at the ONH involving two separate visits within six weeks. Vascular images are generated with OMAG processing by detecting the differences in OCT signals between consecutive B-scans acquired at the same retina location. ONH perfusion is quantified as flux, vessel area density, and normalized flux within the ONH for the prelaminar, lamina cribrosa, and the full ONH. Coefficient of variation (CV) and intraclass correlation coefficient (ICC) are used to evaluate intravisit and intervisit repeatability, and interobserver reproducibility. ONH perfusion measurements show high repeatability [CV≤3.7% (intravisit) and ≤5.2% (intervisit)] and interobserver reproducibility (ICC≤0.966) in all three layers by three metrics. OCTA provides a noninvasive method to visualize and quantify ONH perfusion in human eyes with excellent repeatability and reproducibility, which may add additional insight into ONH perfusion in clinical practice.

Optic nerve head perfusion in normal eyes and eyes with glaucoma using optical coherence tomography-based microangiography

BACKGROUND

To investigate the differences of perfusion in the optic nerve head (ONH) between normal and glaucomatous eyes using optical microangiography (OMAG) based optical coherence tomography (OCT) angiography technique.

METHODS

One eye from each subject was scanned with a 68 kHz Cirrus 5000 HD-OCT-based OMAG prototype system centered at the ONH (Carl Zeiss Meditec Inc, Dublin, CA, USA). Microvascular images were generated from the OMAG dataset by detecting the differences in OCT signal between consecutive B-scans. The pre-laminar layer (preLC) was isolated by a semi-automatic segmentation program. En face OMAG images for preLC were generated using signals with highest blood flow signal intensity. ONH perfusion was quantified as flux, vessel area density, and normalized flux within the ONH. Standard t-tests were performed to analyze the ONH perfusion differences between normal and glaucomatous eyes. Linear regression models were constructed to analyze the correlation between ONH perfusion and other clinical measurements.

RESULTS

Twenty normal and 21 glaucoma subjects were enrolled. Glaucomatous eyes had significantly lower ONH perfusion in preLC in all three perfusion metrics compared to normal eyes (P≤0.0003). Significant correlations between ONH perfusion and disease severity as well as structural changes were detected in glaucomatous eyes (P≤0.012).

CONCLUSIONS

ONH perfusion detected by OMAG showed significant differences between glaucoma and normal controls and was significantly correlated with disease severity and structural defects in glaucomatous eyes. ONH perfusion measurement using OMAG may provide useful information for detection and monitoring of glaucoma.

Successful Implementation of a Program for Increasing Donor Eyes for Research: The Duke-Miracles In Sight Program

PURPOSE

Although essential for development of ocular therapeutics, the quality and quantity of human donor eyes for research have been on the decline. To streamline procurement protocols, provide better medical documentation of tissue, and improve freshness and number of eyes, a pilot cooperative program was undertaken between the Duke University Eye Center and Miracles In Sight Eye Bank.

METHODS

For research eye donors who expire at Duke University Hospital, age restrictions to procurement were lifted, access to donors' electronic medical records was granted to researchers, and eye tissue was delivered directly to scientists. The number of eye pairs received per month and the time from death to arrival in the laboratory were documented, and independent-samples t-tests were used to compare the number of paired eyes and the death-to-laboratory time before and after implementation of the program. A cost analysis of the program was also conducted.

RESULTS

Implementation of the program decreased the time from death to arrival in the laboratory from an average of 22.1 ± 1.5 h (n = 22) to 11.6 ± 0.8 h (n = 75) for a pair of eyes (P < 0.0001). Moreover, the number of whole eye pairs increased from 1.57 ± 0.32 to 3.26 ± 0.27 donors per month (P = 0.0019). Cost analysis indicates that our program is financially viable and sustainable for the eye bank.

CONCLUSIONS

The Duke-Miracles In Sight Program implemented a number of operational changes that resulted in improved quantity and quality of ocular tissue to researchers. Such a model appears feasible for adoption between other eye centers and eye banks.

Optic nerve head perfusion in normal eyes and eyes with glaucoma using optical coherence tomography-based microangiography

BACKGROUND

To investigate the differences of perfusion in the optic nerve head (ONH) between normal and glaucomatous eyes using optical microangiography (OMAG) based optical coherence tomography (OCT) angiography technique.

METHODS

One eye from each subject was scanned with a 68 kHz Cirrus 5000 HD-OCT-based OMAG prototype system centered at the ONH (Carl Zeiss Meditec Inc, Dublin, CA, USA). Microvascular images were generated from the OMAG dataset by detecting the differences in OCT signal between consecutive B-scans. The pre-laminar layer (preLC) was isolated by a semi-automatic segmentation program. En face OMAG images for preLC were generated using signals with highest blood flow signal intensity. ONH perfusion was quantified as flux, vessel area density, and normalized flux within the ONH. Standard t-tests were performed to analyze the ONH perfusion differences between normal and glaucomatous eyes. Linear regression models were constructed to analyze the correlation between ONH perfusion and other clinical measurements.

RESULTS

Twenty normal and 21 glaucoma subjects were enrolled. Glaucomatous eyes had significantly lower ONH perfusion in preLC in all three perfusion metrics compared to normal eyes (P≤0.0003). Significant correlations between ONH perfusion and disease severity as well as structural changes were detected in glaucomatous eyes (P≤0.012).

CONCLUSIONS

ONH perfusion detected by OMAG showed significant differences between glaucoma and normal controls and was significantly correlated with disease severity and structural defects in glaucomatous eyes. ONH perfusion measurement using OMAG may provide useful information for detection and monitoring of glaucoma.

Radiation-related cancer risk associated with surveillance imaging for metastasis from choroidal melanoma

OBJECTIVE

To estimate the lifetime attributable risk of cancer associated with whole-body positron emission tomography (PET)/computed tomography (CT) and with CT of the chest, abdomen, and pelvis if performed at various frequencies and for different durations for surveillance of patients with primary choroidal or ciliary body melanoma for distant metastasis.

METHODS

Effective radiation doses for whole-body CT and for CT of the chest, abdomen, and pelvis were calculated using Monte Carlo simulation studies. The effective dose of the PET scan was estimated by multiplying fludeoxyglucose F18 radioactivity with dose coefficients. Lifetime attributable risks of cancer were calculated using the approach described in the Biological Effects of Ionizing Radiation VII report.

RESULTS

For a 50-year-old patient, an annual CT of the chest, abdomen, and pelvis for 10 years carries an estimated lifetime attributable risk of cancer of 0.9% for male patients and 1.3% for female patients, whereas an annual PET/CT each year for 10 years carries an estimated lifetime attributable risk of cancer of 1.6% for male patients and 1.9% for female patients. Lifetime risk was found to be higher in younger, female patients. The lifetime attributable risk of cancer was estimated to be as high as 7.9% for a 20-year-old female patient receiving a PET/CT scan every 6 months for 10 years.

CONCLUSIONS

Aggressive surveillance protocols incorporating CT scanning or PET/CT scanning for detection of metastasis from primary choroidal or ciliary body melanoma appear to confer a significant substantial risk of a secondary malignant tumor in patients who do not succumb to metastatic melanoma within the first few posttreatment years.

Bacterial dispersal associated with speech in the setting of intravitreous injections

OBJECTIVE

To investigate the amount of bacterial dispersal associated with speech in a simulated intravitreous injection.

METHODS

Fifteen volunteers were recruited. Each volunteer was positioned over an open blood agar plate and did the following: read a 5-minute script with a face mask, read a 5-minute script without a face mask, read a 5-minute script with the face turned away from the plate without a face mask, and stood in silence for 5 minutes. Each volunteer then read a 5-minute script while reclined in a standard ophthalmic examination chair with an open blood agar plate secured to the forehead to simulate bacterial dispersal associated with a talking patient. Total numbers of colony-forming bacteria per plate were counted, and the bacteria were identified.

RESULTS

Significantly less bacterial growth occurred in the face mask and silence conditions compared with the no face mask condition (both P < .001). Bacterial growth was significantly greater in the reclined condition compared with the room control (P = .02). Oral streptococcal species represented 66.7% to 82.6% of bacterial colonies in the no face mask, face turned, and reclined conditions.

CONCLUSIONS

During simulated intravitreous injection, wearing a face mask or remaining silent significantly decreases culture plate contamination from talking. Talking from above and talking in the reclined position were associated with a significant increase in culture plate contamination. Physicians performing intravitreous injections should be aware of these patterns of bacterial contamination, should consider either wearing a face mask or minimizing speech, and should encourage patients to minimize speech during the procedure.

Synthetic lethal analysis of Caenorhabditis elegans posterior embryonic patterning genes identifies conserved genetic interactions

Phenotypic robustness is evidenced when single-gene mutations do not result in an obvious phenotype. It has been suggested that such phenotypic stability results from 'buffering' activities of homologous genes as well as non-homologous genes acting in parallel pathways. One approach to characterizing mechanisms of phenotypic robustness is to identify genetic interactions, specifically, double mutants where buffering is compromised. To identify interactions among genes implicated in posterior patterning of the Caenorhabditis elegans embryo, we measured synthetic lethality following RNA interference of 22 genes in 15 mutant strains. A pair of homologous T-box transcription factors (tbx-8 and tbx-9) is found to interact in both C. elegans and C. briggsae, indicating that their compensatory function is conserved. Furthermore, a muscle module is defined by transitive interactions between the MyoD homolog hlh-1, another basic helix-loop-helix transcription factor, hnd-1, and the MADS-box transcription factor unc-120. Genetic interactions within a homologous set of genes involved in vertebrate myogenesis indicate broad conservation of the muscle module and suggest that other genetic modules identified in C. elegans will be conserved.

The homeodomain protein PAL-1 specifies a lineage-specific regulatory network in the C. elegans embryo

Maternal and zygotic activities of the homeodomain protein PAL-1 specify the identity and maintain the development of the multipotent C blastomere lineage in the C. elegans embryo. To identify PAL-1 regulatory target genes, we used microarrays to compare transcript abundance in wild-type embryos with mutant embryos lacking a C blastomere and to mutant embryos with extra C blastomeres. pal-1-dependent C-lineage expression was verified for select candidate target genes by reporter gene analysis, though many of the target genes are expressed in additional lineages as well. The set of validated target genes includes 12 transcription factors, an uncharacterized wingless ligand and five uncharacterized genes. Phenotypic analysis demonstrates that the identified PAL-1 target genes affect specification, differentiation and morphogenesis of C-lineage cells. In particular, we show that cell fate-specific genes (or tissue identity genes) and a posterior HOX gene are activated in lineage-specific fashion. Transcription of targets is initiated in four temporal phases, which together with their spatial expression patterns leads to a model of the regulatory network specified by PAL-1.

Photoperiod affects neuronal nitric oxide synthase and aggressive behaviour in male Siberian hamsters (Phodopus sungorus)

Many nontropical animals display physiological and behavioural changes in response to seasonal environmental cues including photoperiod (day length). Male Siberian hamsters (Phodopus sungorus) housed in short photoperiod undergo testicular regression accompanied by reduced circulating testosterone and decreased reproductive behaviour. By contrast to the majority of small mammals studied, aggressive behaviour is elevated in short-day Siberian hamsters when blood testosterone concentrations are not detectable. Because gonadal steroid hormones influence neuronal nitric oxide synthase (nNOS), and this enzyme has been implicated in aggressive behaviour, we hypothesized that nNOS expression would be decreased in short-day male Siberian hamsters and negatively correlated with the display of territorial aggression. Adult male Siberian hamsters were individually housed in either long (LD 16:8 h) or short (LD 8:16 h) photoperiods for 10 weeks. Hamsters were assigned to one of two categories by assessing testicular volume and plasma testosterone values: (i) photoperiodic responsive (i.e. regressed testes and low testosterone concentrations) or (ii) photoperiodic nonresponsive (i.e. testes size and circulating testosterone concentrations equivalent to hamsters maintained in long days). At week 10, aggression was assessed using a resident-intruder test. Latency to initial attack, frequency of attacks and duration of total attacks were recorded during a 10-min aggression trial. Brains were collected immediately after behavioural testing and stained for nNOS expression using immunohistochemistry. All short day-housed hamsters were significantly more aggressive than long-day animals, regardless of gonadal size or testosterone concentrations. Short-day animals, both reproductively responsive and nonresponsive morphs, also had significantly less nNOS-immunoreactive cells in the anterior and basolateral amygdaloid areas and paraventricular nuclei compared to long-day hamsters. Together, these results suggest that seasonal aggression in male Siberian hamsters is regulated by photoperiod, through mechanisms that are likely independent from gonadal steroid hormones.