Development and validation of a deep-learning model to predict 10-year atherosclerotic cardiovascular disease risk from retinal images using the UK Biobank and EyePACS 10K datasets.

Background: Atherosclerotic cardiovascular disease (ASCVD) is a leading cause of death globally, and early detection of high-risk individuals is essential for initiating timely interventions. The authors aimed to develop and validate a deep learning (DL) model to predict an individual's elevated 10-year ASCVD risk score based on retinal images and limited demographic data. Methods: The study used 89,894 retinal fundus images from 44,176 UK Biobank participants (96% non-Hispanic White, 5% diabetic) to train and test the DL model. The DL model was developed using retinal images plus age, race/ethnicity, and sex at birth to predict an individual's 10-year ASCVD risk score using the pooled cohort equation (PCE) as the ground truth. This model was then tested on the US EyePACS 10K dataset (5.8% non-Hispanic White, 99.9% diabetic), composed of 18,900 images from 8969 diabetic individuals. Elevated ASCVD risk was defined as a PCE score of ≥7.5%. Results: In the UK Biobank internal validation dataset, the DL model achieved an area under the receiver operating characteristic curve of 0.89, sensitivity 84%, and specificity 90%, for detecting individuals with elevated ASCVD risk scores. In the EyePACS 10K and with the addition of a regression-derived diabetes modifier, it achieved sensitivity 94%, specificity 72%, mean error -0.2%, and mean absolute error 3.1%. Conclusion: This study demonstrates that DL models using retinal images can provide an additional approach to estimating ASCVD risk, as well as the value of applying DL models to different external datasets and opportunities about ASCVD risk assessment in patients living with diabetes.

Alterations in Lens Free Water Distribution Are Associated with Shape Deformation in Accommodation.

Objective: To investigate whether a redistribution of water within the crystalline lens is associated with the shape deformation that occurs during accommodation. Design: Observational, cross sectional study. Subjects: Eleven young adults without presbyopia (aged 18-39 years) and 9 middle-aged adults with presbyopia (aged 40-55 years). Methods: Magnetic resonance imaging (MRI) scans of the lens were acquired on a 3 Tesla clinical MRI scanner, without and with the presentation of a 3 Diopter accommodative stimulus. The MRIs were postprocessed using established methods to extract the geometric dimensions and spatial maps of water distribution of the lens. Main Outcome Measures: Accommodative changes in the full 3-dimensional description of lens shape, the lens total-water distribution profile, and the lens free-water distribution profile. Results: Viewing of an accommodative stimulus by young subjects elicited an elastic shape deformation of the lens consistent with accommodation that was associated with an elevated, smoother free-water distribution, primarily in the anterior region of the lens. In contrast, viewing of an accommodative stimulus by presbyopic subjects produced an atypical shape deformation of the lens that was instead associated with a lowered free-water distribution, primarily in the anterior region of the lens. No discernible changes to the lens total-water distribution were observed in response to the accommodative stimulus in either subject cohort. Conclusions: The present study suggests that protein-mediated alterations in the free-water distribution of the anterior region of the lens influence the shape deformation in accommodation, presenting pharmacological modulation of free-water distribution as an attractive novel approach for treating presbyopia. Financial Disclosures: The authors have no proprietary or commercial interest in any materials discussed in this article.

Examination of alternative eGFR definitions on the performance of deep learning models for detection of chronic kidney disease from fundus photographs.

Deep learning (DL) models have shown promise in detecting chronic kidney disease (CKD) from fundus photographs. However, previous studies have utilized a serum creatinine-only estimated glomerular rate (eGFR) equation to measure kidney function despite the development of more up-to-date methods. In this study, we developed two sets of DL models using fundus images from the UK Biobank to ascertain the effects of using a creatinine and cystatin-C eGFR equation over the baseline creatinine-only eGFR equation on fundus image-based DL CKD predictors. Our results show that a creatinine and cystatin-C eGFR significantly improved classification performance over the baseline creatinine-only eGFR when the models were evaluated conventionally. However, these differences were no longer significant when the models were assessed on clinical labels based on ICD10. Furthermore, we also observed variations in model performance and systemic condition incidence between our study and the ones conducted previously. We hypothesize that limitations in existing eGFR equations and the paucity of retinal features uniquely indicative of CKD may contribute to these inconsistencies. These findings emphasize the need for developing more transparent models to facilitate a better understanding of the mechanisms underpinning the ability of DL models to detect CKD from fundus images.

Automation of Macular Degeneration Classification in the AREDS Dataset, Using a Novel Neural Network Design.

Purpose: To create an ensemble of Convolutional Neural Networks (CNNs), capable of detecting and stratifying the risk of progressive age-related macular degeneration (AMD) from retinal photographs. Design: Retrospective cohort study. Methods: Three individual CNNs are trained to accurately detect 1) advanced AMD, 2) drusen size and 3) the presence or otherwise of pigmentary abnormalities, from macular centered retinal images were developed. The CNNs were then arranged in a "cascading" architecture to calculate the Age-related Eye Disease Study (AREDS) Simplified 5-level risk Severity score (Risk Score 0 - Risk Score 4), for test images. The process was repeated creating a simplified binary "low risk" (Scores 0-2) and "high risk" (Risk Score 3-4) classification. Participants: There were a total of 188,006 images, of which 118,254 images were deemed gradable, representing 4591 patients, from the AREDS1 dataset. The gradable images were split into 50%/25%/25% ratios for training, validation and test purposes. Main Outcome Measures: The ability of the ensemble of CNNs using retinal images to predict an individual's risk of experiencing progression of their AMD based on the AREDS 5-step Simplified Severity Scale. Results: When assessed against the 5-step Simplified Severity Scale, the results generated by the ensemble of CNN's achieved an accuracy of 80.43% (quadratic kappa 0.870). When assessed against a simplified binary (Low Risk/High Risk) classification, an accuracy of 98.08%, sensitivity of ≥85% and specificity of ≥99% was achieved. Conclusion: We have created an ensemble of neural networks, trained on the AREDS 1 dataset, that is able to accurately calculate an individual's score on the AREDS 5-step Simplified Severity Scale for AMD. If the results presented were replicated, then this ensemble of CNNs could be used as a screening tool that has the potential to significantly improve health outcomes by identifying asymptomatic individuals who would benefit from AREDS2 macular supplements.

A multi-centre prospective evaluation of THEIA™ to detect diabetic retinopathy (DR) and diabetic macular oedema (DMO) in the New Zealand screening program.

PURPOSE: To validate the potential application of THEIA™ as clinical decision making assistant in a national screening program. METHODS: A total of 900 patients were recruited from either an urban large eye hospital, or a semi-rural optometrist led screening provider, as they were attending their appointment as part of New Zealand Diabetic Eye Screening Programme. The de-identified images were independently graded by three senior specialists, and final results were aggregated using New Zealand grading scheme, which was then converted to referable/non-referable and Healthy/mild/more than mild/sight threatening categories. RESULTS: THEIA™ managed to grade all images obtained during the study. Comparing the adjudicated images from the specialist grading team, "ground truth", with the grading by the AI platform in detecting "sight threatening" disease, at the patient level THEIA™ achieved 100% imageability, 100% [98.49-100.00%] sensitivity and [97.02-99.16%] specificity, and negative predictive value of 100%. In other words, THEIA™ did not miss any patients with "more than mild" or "sight threatening" disease. The level of agreement between the clinicians and the aggregated results was (k value: 0.9881, 0.9557, and 0.9175), and the level of agreement between THEIA™ and the aggregated labels was (k value: 0.9515). CONCLUSION: This multi-centre prospective trial showed that THEIA™ did not miss referable disease when screening for diabetic retinopathy and maculopathy. It also had a very high level of granularity in reporting the disease level. As THEIA™ has been tested on a variety of cameras, operating in a range of clinics (rural/urban, ophthalmologist-led\optometrist-led), we believe that it will be a suitable addition to a public diabetic screening program.

Patients Perceptions of Artificial Intelligence in Diabetic Eye Screening.

PURPOSE: Artificial intelligence (AI) technology is poised to revolutionize modern delivery of health care services. We set to evaluate the patient perspective of AI use in diabetic retinal screening. DESIGN: Survey. METHODS: Four hundred thirty-eight patients undergoing diabetic retinal screening across New Zealand participated in a survey about their opinion of AI technology in retinal screening. The survey consisted of 13 questions covering topics of awareness, trust, and receptivity toward AI systems. RESULTS: The mean age was 59 years. The majority of participants identified as New Zealand European (50%), followed by Asian (31%), Pacific Islander (10%), and Maori (5%). Whilst 73% of participants were aware of AI, only 58% have heard of it being implemented in health care. Overall, 78% of respondents were comfortable with AI use in their care, with 53% saying they would trust an AI-assisted screening program as much as a health professional. Despite having a higher awareness of AI, younger participants had lower trust in AI systems. A higher proportion of Maori and Pacific participants indicated a preference toward human-led screening. The main perceived benefits of AI included faster diagnostic speeds and greater accuracy. CONCLUSIONS: There is low awareness of clinical AI applications among our participants. Despite this, most are receptive toward the implementation of AI in diabetic eye screening. Overall, there was a strong preference toward continual involvement of clinicians in the screening process. There are key recommendations to enhance the receptivity of the public toward incorporation of AI into retinal screening programs.

Inter-day repeatability assessment of human retinal blood flow using clinical laser speckle contrast imaging.

Laser speckle contrast imaging (LSCI) can generate retinal blood flow maps inexpensively and non-invasively. These flow maps can be used to identify various eye disorders associated with reduced blood flow. Despite early success, one of the major obstacles to clinical adoption of LSCI is poor repeatability of the modality. Here, we propose an LSCI registration pipeline that registers contrast maps to correct for rigid movements. Post-registration, intra(same)-day and inter(next)-day repeatability are studied using various quantitative metrics. We have studied LSCI repeatability intra-day by using the coefficient of variation. Using the processing pipelines and custom hardware developed, similar repeatability was observed when compared to previously reported values in the literature. Inter-day repeatability analysis indicates no statistical evidence (p = 0.09) of a difference between flow measurements performed on two independent days. Further improvements to hardware, environmental controls, and participant control must be made to provide higher confidence in the repeatability of blood flow. However, this is the first time that repeatability across two different days (inter-day) using multiple exposure speckle imaging (MESI) has been analyzed and reported.

Evaluation of the prevalence of non-diabetic eye disease detected at first screen from a single region diabetic retinopathy screening program: a cross-sectional cohort study in Auckland, New Zealand.

OBJECTIVES: To evaluate the prevalence of incidental non-diabetic ocular comorbidities detected at first screen in a large diabetic retinopathy (DR) screening programme. DESIGN: Cross-sectional cohort study. SETTING: Single large metropolitan diabetic eye screening programme in Auckland, New Zealand. PARTICIPANTS: Twenty-two thousand seven hundred and seventy-one participants who attended screening from September 2008 to August 2018. RESULTS: Hypertensive retinopathy (HTR) was observed in 14.2% (3236/22 771) participants. Drusen were present in 14.0% participants under the age of 55 years, increasing to 20.5% in those 55 years and older. The prevalence of neovascular age-related macular degeneration (AMD) was 0.5% in participants aged<55 years, 2.4% in participants aged 55-75 years and 16% in participants aged>75 years. Retinal vein occlusion and retinal arterial embolus were prevalent in 0.7% and 0.02%, respectively, in participants aged<55 years, increasing to 2.2% and 0.4%, respectively, in those >75 years. Cataracts were common being present in 37.1% of participants over the age of 75 years. Only 386 individuals (1.7%) were labelled as glaucoma suspects. Geographic atrophy, epiretinal membrane, choroidal nevi and posterior capsular opacification had an increased prevalence in older individuals. CONCLUSIONS: Our data suggest that AMD, HTR and cataracts are routinely detected during DR screening. The incorporation of the detection of these ocular comorbidities during DR screening provide opportunities for patients to modify risk factors (smoking cessation and diet for AMD, blood pressure for HTR) and allow access to cataract surgery.

Risk Factors for Progression to Referable Diabetic Eye Disease in People With Diabetes Mellitus in Auckland, New Zealand: A 12-Year Retrospective Cohort Analysis.

PURPOSE: To evaluate the prevalence and risk factors for the development of any and referable diabetic eye disease in a multi-ethnic New Zealand population with diabetes mellitus attending a regional retinal screening service. METHODS: Retrospective observational cohort study of people living with diabetes who attended the Auckland Regional Diabetic Retinal Screening Programme 2006-2018 inclusive (n = 41,786). RESULTS: Any retinopathy/maculopathy was present at first screening for 48.2% [95% confidence interval (CI): 45.8%-50.6%] / 37.8% (95% CI: 35.5%- 40.1%) of people with Type 1 and 25% (95% CI: 24.6%-25.4%) / 21.9% (95% CI: 21.5%-22.3%) with Type 2 diabetes. Referable retinopathy at baseline screening was 4.4% (95% CI: 3.6%-5.3%) and 1.6% (95% CI: 1.5%-1.7%) among people with Type 1 and Type 2 diabetes mellitus, respectively. After 4 years, cumulative incidence for referable retinopathy /referable maculopathy was 12/36 per 1000 people with Type 1 and 2.4/16 per 1000 people with Type 2 diabetes. Independent hazards for disease progression varied for the diabetes cohort types but baseline grade, duration of diabetes, and HbA1c were common to all. CONCLUSIONS: Referable diabetic eye disease at the first screening and after 4 years of follow-up is uncommon. Lengthening of the screening intervals for people with no or mild diabetic eye disease at first screening assessment could be considered.

Hyperbaric oxygen as a model of lens aging in the bovine lens: The effects on lens biochemistry, physiology and optics.

Age related nuclear (ARN) cataracts in humans take years to form and so experimental models have been developed to mimic the process in animals as a means of better understanding the etiology of nuclear cataracts in humans. A major limitation with these animal models is that many of the biochemical and physiological changes are not typical of that seen in human ARN cataract. In this review, we highlight the work of Frank Giblin and colleagues who established an in vivo animal model that replicates many of the changes observed in human ARN cataract. This model involves exposing aged guinea pigs to hyperbaric oxygen (HBO), which by causing the depletion of the antioxidant glutathione (GSH) specifically in the lens nucleus, produces oxidative changes to nuclear proteins, nuclear light scattering and a myopic shift in lens power that mimics the change that often precedes cataract development in humans. However, this model involves multiple HBO treatments per week, with sometimes up to a total of 100 treatments, spanning up to eight months, which is both costly and time consuming. To address these issues, Giblin developed an in vitro model that used rabbit lenses exposed to HBO for several hours which was subsequently shown to replicate many of the changes observed in human ARN cataract. These experiments suggest that HBO treatment of in vitro animal lenses may serve as a more economical and efficient model to study the development of cataract. Inspired by these experiments, we investigated whether exposure of young bovine lenses to HBO for 15 h could also serve as a suitable acute model of ARN cataract. We found that while this model is able to exhibit some of the biochemical and physiological changes associated with ARN cataract, the decrease in lens power we observed was more characteristic of the hyperopic shift in refraction associated with ageing. Future work will investigate whether HBO treatment to age the bovine lens in combination with an oxidative stressor such as UV light will induce refractive changes more closely associated with human ARN cataract. This will be important as developing an animal model that replicates the changes to lens biochemistry, physiology and optics observed in human ARN cataracts is urgently required to facilitate the identification and testing of anti-cataract therapies that are effective in humans.

Age-Dependent Changes in Total and Free Water Content of In Vivo Human Lenses Measured by Magnetic Resonance Imaging.

Purpose: To use magnetic resonance imaging (MRI) to measure age-dependent changes in total and free water in human lenses in vivo. Methods: Sixty-four healthy adults aged 18 to 86 years were recruited, fitted with a 32-channel head receiver coil, and placed in a 3 Tesla clinical MR scanner. Scans of the crystalline lens were obtained using a volumetric interpolated breath-hold examination sequence with dual flip angles, which were corrected for field inhomogeneity post-acquisition using a B1-map obtained using a turbo-FLASH sequence. The spatial distribution and content of corrected total (ρlens) and free (T1) water along the lens optical axis were extracted using custom-written code. Results: Lens total water distribution and content did not change with age (all P > 0.05). In contrast to total water, a gradient in free water content that was highest in the periphery relative to the center was present in lenses across all ages. However, this initially parabolic free water gradient gradually developed an enhanced central plateau, as indicated by increasing profile shape parameter values (anterior: 0.067/y, P = 0.004; posterior: 0.050/y, P = 0.020) and central free water content (1.932 ms/y, P = 0.022) with age. Conclusions: MRI can obtain repeatable total and free water measurements of in vivo human lenses. The observation that the lens steady-state free, but not total, water gradient is abolished with age raises the possibility that alterations in protein-water interactions are an underlying cause of the degradation in lens optics and overall vision observed with aging.

Intraocular pressure fluctuation during resistance exercise.

OBJECTIVE: To evaluate the effect of weightlifting (leg press) on intraocular pressure (IOP). DESIGN: Prospective cohort study. SUBJECTS: A total of 24 participants met the inclusion criteria and completed the study procedures. Participants had an average age of 22.7±2.7 years and included nine women. The mean baseline IOP was 13.9 mm Hg (SD=2.4) with an average body mass index of 24.5 (SD= 3.1). METHODS: The maximum load for a single lift was found for each participant. Participants then performed three leg press regimens: one repetition using 95% of maximal load (1RM), six repetitions using 75% of maximal load (6RM) and isometric push against a weight much heavier than maximal load (ISO). MAIN OUTCOME MEASURE: IOP was measured pre-exercise, during and immediately following the exercise using an iCare TA01i rebound tonometer. Blood pressure and HR were being monitored continuously during the lift. Optical coherence tomography images were obtained pre and postexercise session. RESULTS: The average maximum weight lifted by our participants was 331.9 Kg (SD=97.3). Transient increased IOP was observed across the 1RM, 6RM and ISO exercises with an average increase in 26.4 mm Hg (23.7 mm Hg to 28.7 mm Hg) to reach an average max IOP of 40.7 mm Hg (27.8 mm Hg to 54.2 mm Hg), with an absolute maximum of 70 mm Hg in one participant. CONCLUSIONS: There is a transient and dramatic fluctuation in IOP with resistance training. This coupled with regular exposure to resistance training is potentially a significant risk factor for glaucoma. It should be noted that this study has been carried out in a healthy young population, and, thus, the external validity of these results in glaucoma participants requires further investigation.

Using the Lens Paradox to Optimize an In Vivo MRI-Based Optical Model of the Aging Human Crystalline Lens.

Purpose: To optimize our in vivo magnetic resonance imaging (MRI)-based optical model of the human crystalline lens, developed with a small group of young adults, for a larger cohort spanning a wider age range. Methods: Subjective refraction and ocular biometry were measured in 57 healthy adults ages 18 to 86 years who were then scanned using 3T clinical magnetic resonance imaging (MRI) to obtain lens gradient of refractive index (GRIN) and geometry measurements. These parameters were combined with ocular biometric measurements to construct individualized Zemax eye models from which ocular refractive errors and lens powers were determined. Models were optimized by adding an age-dependent factor to the transverse relaxation time (T2)-refractive index (n) calibration to match model-calculated refractive errors with subjective refractions. Results: In our subject cohort, subjective refraction shifted toward hyperopia by 0.029 diopter/year as the lens grew larger and developed flatter GRINs with advancing age. Without model optimization, lens powers did not reproduce this clinically observed decrease, the so-called lens paradox, instead increasing by 0.055 diopter/year. However, modifying the T2-n calibration by including an age-dependent factor reproduced the decrease in lens power associated with the lens paradox. Conclusions: After accounting for age-related changes in lens physiology in the T2-n calibration, our model was capable of accurately measuring in vivo lens power across a wide age range. This study highlights the need for a better understanding of how age-dependent changes to the GRIN impact the refractive properties of the lens. Translational Relevance: MRI is applied clinically to calculate the effect of age-related refractive index changes in the lens paradox.

Multi-parametric MRI of the physiology and optics of the in-vivo mouse lens.

The optics of the ocular lens are determined by its geometry (shape and volume) and its inherent gradient of refractive index (water to protein ratio), which are in turn maintained by unique cellular physiology known as the lens internal microcirculation system. Previously, magnetic resonance imaging (MRI) has been used on ex vivo organ cultured bovine lenses to show that pharmacological perturbations to this microcirculation system disrupt ionic and fluid homeostasis and overall lens optics. In this study, we have optimised in vivo MRI protocols for use on wild-type and transgenic mouse models so that the effects of genetically perturbing the lens microcirculation system on lens properties can be studied. In vivo MRI protocols and post-analysis methods for studying the mouse lens were optimised and used to measure the lens geometry, diffusion, T1 and T2, as well as the refractive index (n) calculated from T2, in wild-type mice and the genetically modified Cx50KI46 mouse. In this animal line, gap junctional coupling in the lens is increased by knocking in the gap junction protein Cx46 into the Cx50 locus. Relative to wild-type mice, Cx50KI46 mice showed significantly reduced lens size and radius of curvature, increased T1 and T2 values, and decreased n in the lens nucleus, which was consistent with the developmental and functional changes characterised previously in this lens model. These proof of principle experiments show that in vivo MRI can be applied to transgenic mouse models to gain mechanistic insights into the relationship between lens physiology and optics, and in the future suggest that longitudinal studies can be performed to determine how this relationship is altered by age in mouse models of cataract.

Multimodal Retinal Image Analysis via Deep Learning for the Diagnosis of Intermediate Dry Age-Related Macular Degeneration: A Feasibility Study.

RESULTS: The CNN trained using OCT alone showed a diagnostic accuracy of 94%, whilst the OCT-A trained CNN resulted in an accuracy of 91%. When multiple modalities were combined, the CNN accuracy increased to 96% in the AMD cohort. CONCLUSIONS: Here we demonstrate that superior diagnostic accuracy can be achieved when deep learning is combined with multimodal image analysis.

Towards implementation of AI in New Zealand national diabetic screening program: Cloud-based, robust, and bespoke.

Convolutional Neural Networks (CNNs) have become a prominent method of AI implementation in medical classification tasks. Grading Diabetic Retinopathy (DR) has been at the forefront of the development of AI for ophthalmology. However, major obstacles remain in the generalization of these CNNs onto real-world DR screening programs. We believe these difficulties are due to use of 1) small training datasets (<5,000 images), 2) private and 'curated' repositories, 3) locally implemented CNN implementation methods, while 4) relying on measured Area Under the Curve (AUC) as the sole measure of CNN performance. To address these issues, the public EyePACS Kaggle Diabetic Retinopathy dataset was uploaded onto Microsoft Azure™ cloud platform. Two CNNs were trained; 1 a "Quality Assurance", and 2. a "Classifier". The Diabetic Retinopathy classifier CNN (DRCNN) performance was then tested both on 'un-curated' as well as the 'curated' test set created by the "Quality Assessment" CNN model. Finally, the sensitivity of the DRCNNs was boosted using two post-training techniques. Our DRCNN proved to be robust, as its performance was similar on 'curated' and 'un-curated' test sets. The implementation of 'cascading thresholds' and 'max margin' techniques led to significant improvements in the DRCNN's sensitivity, while also enhancing the specificity of other grades.

Relationship between rheological properties and transverse relaxation time (T2) of artificial and porcine vitreous humour.

Vitreous liquefactive processes play an integral role in ocular health. Knowledge of the degree of liquefaction would allow better monitoring of ocular disease progression and enable more informed therapeutic dosing for an individual patient. Presently this process cannot be monitored in a non-invasive manner. Here, we evaluated whether magnetic resonance imaging (MRI) could predict the viscoelasticity and in turn liquefactive state of artificial and biological vitreous humour. Gels comprising identical concentrations of hyaluronic acid and agar ranging from 0.125 to 2.25 mg/ml of each polymer were prepared and their T2 was measured using a turbo-spin echo sequence via 3T clinical MRI. The gels were subsequently subjected to rheological frequency and flow sweeps and trends between T2 and rheological parameters were assessed. The relationship between T2 and vitreous humour rheology was further assessed using ex vivo porcine eyes. An optimised imaging technique improved homogeneity of obtained artificial vitreous humour T2 maps. Strong correlations were observed between T2 and various rheological parameters of the gels. Translation to porcine vitreous humour demonstrated that the T2 of biological tissue was related to its viscoelastic properties. This study shows that T2 can be correlated with various rheological parameters within gels. Future investigations will assess the translatability of these findings to live models.

Quantification of Optical Coherence Tomography Angiography in Age and Age-Related Macular Degeneration Using Vessel Density Analysis.

PURPOSE: The aim of this study was to determine whether vessel density (VD) as measured by optical coherence tomography (OCT) angiography provided insights into retinal and choriocapillaris vascular changes with aging and intermediate dry age-related macular degeneration (AMD). DESIGN: Non-randomized observational study. METHODS: Seventy-five participants were recruited into 3 cohorts: young healthy group, old healthy, and those at high-risk for exudative AMD. Raw OCT and OCT angiography data from TOPCON DRI OCT Triton were exported using Topcon IMAGENET 6.0 software, and 3D datasets were analysed to determine retinal thickness and VD. RESULTS: Central macular thickness measurements revealed a trend of overall retinal thinning with increasing age. VD through the full thickness of the retina was highest in Early Treatment Diabetic Retinopathy Study (ETDRS) sector 4 (the inferior macula) in all the cohorts. Mean VD was significantly higher in the deep capillary plexus than the superficial capillary plexus in all ETDRS sectors in all cohorts, but there was no significant difference noted between groups. Choriocapillaris VD was significantly lower in all ETDRS sectors in the AMD group compared with the young healthy and the old healthy groups. CONCLUSIONS: Retinal VD maps, derived from the retinal plexi, are not reliable biomarkers for assessing the aging macular. Our nonproprietary analysis of the vascular density of the choriocapillaris revealed a significant drop off of VD with age and disease, but further work is required to corroborate this finding. If repeatable, choriocapillaris VD may provide a noninvasive biomarker of healthy aging and disease.

Development of an in vivo magnetic resonance imaging and computer modelling platform to investigate the physiological optics of the crystalline lens.

We have developed and validated in vivo magnetic resonance imaging (MRI) protocols to extract parameters (T2 and geometry) of the human lens that, combined with biometric measures of the eye and optical modelling, enable us to investigate the relative contributions made by the gradient of refractive index (GRIN) and the shape of the lens to the refractive properties of each subject tested. Seven young and healthy participants (mean age: 25.6 ± 3.6 years) underwent an ophthalmic examination, and two sessions of MRI scans using a 3 T clinical magnet. Our MRI protocols for studying lens physiological optics and geometrical measurements were repeatable and reliable, using both 1D (95% confidence interval (CI) for mean differences for exponents = [-2.1, 2.6]) and 2D analysis (anterior T2 CI for differences [-6.4, 8.1] ms; posterior T2 CI for differences [-6.4, 8.3] ms). The lens thickness measured from MRI showed good correlation with that measured with clinical 'gold standard' LenStar (mean differences = [-0.18, 0.2] mm). The predicted refractive errors from ZEMAX had reasonable agreements with participants' clinic records (mean differences = [-1.7, 1.2] D). Quantitative measurements of lens geometry and GRIN with our MRI technique showed high inter-day repeatability. Our clinical MRI technique also provides reliable measures of lens geometry that are comparable to optical biometry. Finally, our ZEMAX optical models produced accurate refractive error and lens power estimations.

Detection of smoking status from retinal images; a Convolutional Neural Network study.

Cardiovascular diseases are directly linked to smoking habits, which has both physiological and anatomical effects on the systemic and retinal circulations, and these changes can be detected with fundus photographs. Here, we aimed to 1- design a Convolutional Neural Network (CNN), using retinal photographs, to differentiate between smokers and non-smokers; and 2- use the attention maps to better understand the physiological changes that occur in the retina in smokers. 165,104 retinal images were obtained from a diabetes screening programme, labelled with self-reported "smoking" or "non-smoking" status. The images were pre-processed in one of two ways, either "contrast-enhanced" or "skeletonized". Experiments were run on an Intel Xeon Gold 6128 CPU @ 3.40 GHz with 16 GB of RAM memory and a NVIDIA GeForce TiTan V VOLTA 12 GB, for 20 epochs. The dataset was split 80/20 for training and testing sets, respectively. The overall validation outcomes for the contrast-enhanced model were accuracy 88.88%, specificity 93.87%. In contrast, the outcomes of the skeletonized model were accuracy 63.63%, specificity 65.60%. The "attention maps" that were generated of the contrast-enhanced model highlighted the retinal vasculature, perivascular region and the fovea most prominently. We trained a customized CNN to accurately determine smoking status. The retinal vasculature, the perivascular region and the fovea appear to be important predictive features in the determination of smoking status. Despite a high degree of accuracy, the sensitivity of our CNN was low. Further research is required to establish whether the frequency, duration, and dosage (quantity) of smoking would improve the sensitivity of the CNN.