A foundation model for generalizable disease detection from retinal images.

Medical artificial intelligence (AI) offers great potential for recognizing signs of health conditions in retinal images and expediting the diagnosis of eye diseases and systemic disorders1. However, the development of AI models requires substantial annotation and models are usually task-specific with limited generalizability to different clinical applications2. Here, we present RETFound, a foundation model for retinal images that learns generalizable representations from unlabelled retinal images and provides a basis for label-efficient model adaptation in several applications. Specifically, RETFound is trained on 1.6 million unlabelled retinal images by means of self-supervised learning and then adapted to disease detection tasks with explicit labels. We show that adapted RETFound consistently outperforms several comparison models in the diagnosis and prognosis of sight-threatening eye diseases, as well as incident prediction of complex systemic disorders such as heart failure and myocardial infarction with fewer labelled data. RETFound provides a generalizable solution to improve model performance and alleviate the annotation workload of experts to enable broad clinical AI applications from retinal imaging.

Leading in the development, standardised evaluation, and adoption of artificial intelligence in clinical practice: regional anaesthesia as an example.

A recent study by Suissa and colleagues explored the clinical relevance of a medical image segmentation metric (Dice metric) commonly used in the field of artificial intelligence (AI). They showed that pixel-wise agreement for physician identification of structures on ultrasound images is variable, and a relatively low Dice metric (0.34) correlated to a substantial agreement on subjective clinical assessment. We highlight the need to bring structure and clinical perspective to the evaluation of medical AI, which clinicians are best placed to direct.

Prevalence of Diabetic Retinopathy in Indigenous and Non-Indigenous Australians: A Systematic Review and Meta-analysis.

TOPIC: This systematic review and meta-analysis summarizes evidence relating to the prevalence of diabetic retinopathy (DR) among Indigenous and non-Indigenous Australians. CLINICAL RELEVANCE: Indigenous Australians suffer disproportionately from diabetes-related complications. Exploring ethnic variation in disease is important for equitable distribution of resources and may lead to identification of ethnic-specific modifiable risk factors. Existing DR prevalence studies comparing Indigenous and non-Indigenous Australians have shown conflicting results. METHODS: This study was conducted following Joanna Briggs Institute guidance on systematic reviews of prevalence studies (PROSPERO ID: CRD42022259048). We performed searches of Medline (Ovid), EMBASE, and Web of Science until October 2021, using a strategy designed by an information specialist. We included studies reporting DR prevalence among diabetic patients in Indigenous and non-Indigenous Australian populations. Two independent reviewers performed quality assessments using a 9-item appraisal tool. Meta-analysis and meta-regression were performed using double arcsine transformation and a random-effects model comparing Indigenous and non-Indigenous subgroups. RESULTS: Fifteen studies with 8219 participants met criteria for inclusion. The Indigenous subgroup scored lower on the appraisal tool than the non-Indigenous subgroup (mean score 50% vs. 72%, P = 0.04). In the unadjusted meta-analysis, DR prevalence in the Indigenous subgroup (30.2%; 95% confidence interval [CI], 24.9-35.7) did not differ significantly (P = 0.17) from the non-Indigenous subgroup (23.7%; 95% CI, 16.8-31.4). After adjusting for age and quality, DR prevalence was higher in the Indigenous subgroup (P < 0.01), with prevalence ratio point estimates ranging from 1.72 to 2.58, depending on the meta-regression model. For the secondary outcomes, prevalence estimates were higher in the Indigenous subgroup for diabetic macular edema (DME) (8.7% vs. 2.7%, P = 0.02) and vision-threatening DR (VTDR) (8.6% vs. 3.0%, P = 0.03) but not for proliferative DR (2.5% vs. 0.8%, P = 0.07). CONCLUSIONS: Indigenous studies scored lower for methodological quality, raising the possibility that systematic differences in research practices may be leading to underestimation of disease burden. After adjusting for age and quality, we found a higher DR prevalence in the Indigenous subgroup. This contrasts with a previous review that reported the opposite finding of lower DR prevalence using unadjusted pooled estimates. Future epidemiological work exploring DR burden in Indigenous communities should aim to address methodological weaknesses identified by this review.

Visual dysfunction is a better predictor than retinal thickness for dementia in Parkinson's disease.

BACKGROUND: Dementia is a common and devastating symptom of Parkinson's disease (PD). Visual function and retinal structure are both emerging as potentially predictive for dementia in Parkinson's but lack longitudinal evidence. METHODS: We prospectively examined higher order vision (skew tolerance and biological motion) and retinal thickness (spectral domain optical coherence tomography) in 100 people with PD and 29 controls, with longitudinal cognitive assessments at baseline, 18 months and 36 months. We examined whether visual and retinal baseline measures predicted longitudinal cognitive scores using linear mixed effects models and whether they predicted onset of dementia, death and frailty using time-to-outcome methods. RESULTS: Patients with PD with poorer baseline visual performance scored lower on a composite cognitive score (ß=0.178, SE=0.05, p=0.0005) and showed greater decreases in cognition over time (ß=0.024, SE=0.001, p=0.013). Poorer visual performance also predicted greater probability of dementia (χ² (1)=5.2, p=0.022) and poor outcomes (χ² (1) =10.0, p=0.002). Baseline retinal thickness of the ganglion cell-inner plexiform layer did not predict cognitive scores or change in cognition with time in PD (ß=-0.013, SE=0.080, p=0.87; ß=0.024, SE=0.001, p=0.12). CONCLUSIONS: In our deeply phenotyped longitudinal cohort, visual dysfunction predicted dementia and poor outcomes in PD. Conversely, retinal thickness had less power to predict dementia. This supports mechanistic models for Parkinson's dementia progression with onset in cortical structures and shows potential for visual tests to enable stratification for clinical trials.

DEEP MOVEMENT: Deep learning of movie files for management of endovascular thrombectomy.

OBJECTIVES: Treatment and outcomes of acute stroke have been revolutionised by mechanical thrombectomy. Deep learning has shown great promise in diagnostics but applications in video and interventional radiology lag behind. We aimed to develop a model that takes as input digital subtraction angiography (DSA) videos and classifies the video according to (1) the presence of large vessel occlusion (LVO), (2) the location of the occlusion, and (3) the efficacy of reperfusion. METHODS: All patients who underwent DSA for anterior circulation acute ischaemic stroke between 2012 and 2019 were included. Consecutive normal studies were included to balance classes. An external validation (EV) dataset was collected from another institution. The trained model was also used on DSA videos post mechanical thrombectomy to assess thrombectomy efficacy. RESULTS: In total, 1024 videos comprising 287 patients were included (44 for EV). Occlusion identification was achieved with 100% sensitivity and 91.67% specificity (EV 91.30% and 81.82%). Accuracy of location classification was 71% for ICA, 84% for M1, and 78% for M2 occlusions (EV 73, 25, and 50%). For post-thrombectomy DSA (n = 194), the model identified successful reperfusion with 100%, 88%, and 35% for ICA, M1, and M2 occlusion (EV 89, 88, and 60%). The model could also perform classification of post-intervention videos as mTICI < 3 with an AUC of 0.71. CONCLUSIONS: Our model can successfully identify normal DSA studies from those with LVO and classify thrombectomy outcome and solve a clinical radiology problem with two temporal elements (dynamic video and pre and post intervention). KEY POINTS: • DEEP MOVEMENT represents a novel application of a model applied to acute stroke imaging to handle two types of temporal complexity, dynamic video and pre and post intervention. • The model takes as an input digital subtraction angiograms of the anterior cerebral circulation and classifies according to (1) the presence or absence of large vessel occlusion, (2) the location of the occlusion, and (3) the efficacy of thrombectomy. • Potential clinical utility lies in providing decision support via rapid interpretation (pre thrombectomy) and automated objective gradation of thrombectomy outcomes (post thrombectomy).

Assistive artificial intelligence for ultrasound image interpretation in regional anaesthesia: an external validation study.

BACKGROUND: Ultrasonound is used to identify anatomical structures during regional anaesthesia and to guide needle insertion and injection of local anaesthetic. ScanNav Anatomy Peripheral Nerve Block (Intelligent Ultrasound, Cardiff, UK) is an artificial intelligence-based device that produces a colour overlay on real-time B-mode ultrasound to highlight anatomical structures of interest. We evaluated the accuracy of the artificial-intelligence colour overlay and its perceived influence on risk of adverse events or block failure. METHODS: Ultrasound-guided regional anaesthesia experts acquired 720 videos from 40 volunteers (across nine anatomical regions) without using the device. The artificial-intelligence colour overlay was subsequently applied. Three more experts independently reviewed each video (with the original unmodified video) to assess accuracy of the colour overlay in relation to key anatomical structures (true positive/negative and false positive/negative) and the potential for highlighting to modify perceived risk of adverse events (needle trauma to nerves, arteries, pleura, and peritoneum) or block failure. RESULTS: The artificial-intelligence models identified the structure of interest in 93.5% of cases (1519/1624), with a false-negative rate of 3.0% (48/1624) and a false-positive rate of 3.5% (57/1624). Highlighting was judged to reduce the risk of unwanted needle trauma to nerves, arteries, pleura, and peritoneum in 62.9-86.4% of cases (302/480 to 345/400), and to increase the risk in 0.0-1.7% (0/160 to 8/480). Risk of block failure was reported to be reduced in 81.3% of scans (585/720) and to be increased in 1.8% (13/720). CONCLUSIONS: Artificial intelligence-based devices can potentially aid image acquisition and interpretation in ultrasound-guided regional anaesthesia. Further studies are necessary to demonstrate their effectiveness in supporting training and clinical practice. CLINICAL TRIAL REGISTRATION: NCT04906018.

Biomarkers of macular neovascularisation activity using optical coherence tomography angiography in treated stable neovascular age related macular degeneration.

BACKGROUND: The aim of this study was to describe features of disease activity in patients with treated stable macular neovascularisation (MNV) in neovascular age related macular degeneration (nAMD) using optical coherence tomography angiography (OCTA). METHODS: Thirty-two eyes of 32 patients with nAMD were included in this prospective, observational study. These patients were undergoing treatment with aflibercept on a treat-and-extend regimen attending an extension to a 12-week treatment interval. RESULTS: All subjects had no macular haemorrhage and no structural OCT markers of active MNV activity at the index 12-week treatment extension visit. 31/32 OCTA images were gradeable without significant imaging artefact. The mean MNV size was 3.6mm2 ± 4.6mm2 and 27 (87.1%) had detectable MNV blood flow. 29/31 (93.5%) subjects had MNV with mature phenotypes including 10 non-specific, 10 tangle and 3 deadtree phenotypes. MNV halo and MNV central feeder vessel were noted in 18 (58.1%) and 19 (61.3%) of subjects respectively; only 1 (3.2%) subject was noted to have a MNV capillary fringe. CONCLUSIONS: MNV blood flow is still detectable using OCTA in the majority of subjects in this study with treated stable MNV. OCTA features associated included MNV mature phenotype, MNV feeder vessel, MNV halo and absence of capillary fringe.

An in vitro assessment of the thermoreversible PLGA-PEG-PLGA copolymer: Implications for Descemet's membrane endothelial keratoplasty.

BACKGROUND: To explore the use of a thermoreversible copolymer gel coating to prevent donor tissue scrolling in Descemet's membrane endothelial keratoplasty (DMEK). METHODS: PLGA-PEG-PLGA triblock copolymer was synthesised via ring opening polymerisation. Two formulations were fabricated and gelation properties characterised using rheological analyses. Endothelial cytotoxicity of the copolymer was assessed using a Trypan Blue exclusion assay. Thickness of the copolymer gel coating on the endothelial surface was analysed using anterior segment optical coherence tomography (OCT) (RTVue-100, Optovue Inc.). Gold nanoparticles were added to the copolymer to aid visualisation using OCT. Prevention of Descemet membrane donor scrolling was represented via a novel, in vitro, immersion of copolymer coated donor graft material. RESULTS: Two different formulations of PLGA-PEG-PLGA copolymer were successfully fabricated and the desired peak gelling temperature of 24°C was achieved by polymer blending. Application of 20%, 30% and 40% (wt/vol) polymer concentrations resulted in a statistically significant increase in polymer thickness on the endothelium (p < 0.001). There was no detectable endothelial cytotoxicity. The polymer was easy to apply to the endothelium and prevented scrolling of the DMEK graft. CONCLUSION: This PLGA-PEG-PLGA thermoreversible copolymer gel could be exploited as a therapeutic aid for preventing DMEK graft scrolling.

Democratizing Artificial Intelligence Imaging Analysis With Automated Machine Learning: Tutorial.

Deep learning-based clinical imaging analysis underlies diagnostic artificial intelligence (AI) models, which can match or even exceed the performance of clinical experts, having the potential to revolutionize clinical practice. A wide variety of automated machine learning (autoML) platforms lower the technical barrier to entry to deep learning, extending AI capabilities to clinicians with limited technical expertise, and even autonomous foundation models such as multimodal large language models. Here, we provide a technical overview of autoML with descriptions of how autoML may be applied in education, research, and clinical practice. Each stage of the process of conducting an autoML project is outlined, with an emphasis on ethical and technical best practices. Specifically, data acquisition, data partitioning, model training, model validation, analysis, and model deployment are considered. The strengths and limitations of available code-free, code-minimal, and code-intensive autoML platforms are considered. AutoML has great potential to democratize AI in medicine, improving AI literacy by enabling "hands-on" education. AutoML may serve as a useful adjunct in research by facilitating rapid testing and benchmarking before significant computational resources are committed. AutoML may also be applied in clinical contexts, provided regulatory requirements are met. The abstraction by autoML of arduous aspects of AI engineering promotes prioritization of data set curation, supporting the transition from conventional model-driven approaches to data-centric development. To fulfill its potential, clinicians must be educated on how to apply these technologies ethically, rigorously, and effectively; this tutorial represents a comprehensive summary of relevant considerations.

Stakeholder Perspectives of Clinical Artificial Intelligence Implementation: Systematic Review of Qualitative Evidence.

BACKGROUND: The rhetoric surrounding clinical artificial intelligence (AI) often exaggerates its effect on real-world care. Limited understanding of the factors that influence its implementation can perpetuate this. OBJECTIVE: In this qualitative systematic review, we aimed to identify key stakeholders, consolidate their perspectives on clinical AI implementation, and characterize the evidence gaps that future qualitative research should target. METHODS: Ovid-MEDLINE, EBSCO-CINAHL, ACM Digital Library, Science Citation Index-Web of Science, and Scopus were searched for primary qualitative studies on individuals' perspectives on any application of clinical AI worldwide (January 2014-April 2021). The definition of clinical AI includes both rule-based and machine learning-enabled or non-rule-based decision support tools. The language of the reports was not an exclusion criterion. Two independent reviewers performed title, abstract, and full-text screening with a third arbiter of disagreement. Two reviewers assigned the Joanna Briggs Institute 10-point checklist for qualitative research scores for each study. A single reviewer extracted free-text data relevant to clinical AI implementation, noting the stakeholders contributing to each excerpt. The best-fit framework synthesis used the Nonadoption, Abandonment, Scale-up, Spread, and Sustainability (NASSS) framework. To validate the data and improve accessibility, coauthors representing each emergent stakeholder group codeveloped summaries of the factors most relevant to their respective groups. RESULTS: The initial search yielded 4437 deduplicated articles, with 111 (2.5%) eligible for inclusion (median Joanna Briggs Institute 10-point checklist for qualitative research score, 8/10). Five distinct stakeholder groups emerged from the data: health care professionals (HCPs), patients, carers and other members of the public, developers, health care managers and leaders, and regulators or policy makers, contributing 1204 (70%), 196 (11.4%), 133 (7.7%), 129 (7.5%), and 59 (3.4%) of 1721 eligible excerpts, respectively. All stakeholder groups independently identified a breadth of implementation factors, with each producing data that were mapped between 17 and 24 of the 27 adapted Nonadoption, Abandonment, Scale-up, Spread, and Sustainability subdomains. Most of the factors that stakeholders found influential in the implementation of rule-based clinical AI also applied to non-rule-based clinical AI, with the exception of intellectual property, regulation, and sociocultural attitudes. CONCLUSIONS: Clinical AI implementation is influenced by many interdependent factors, which are in turn influenced by at least 5 distinct stakeholder groups. This implies that effective research and practice of clinical AI implementation should consider multiple stakeholder perspectives. The current underrepresentation of perspectives from stakeholders other than HCPs in the literature may limit the anticipation and management of the factors that influence successful clinical AI implementation. Future research should not only widen the representation of tools and contexts in qualitative research but also specifically investigate the perspectives of all stakeholder HCPs and emerging aspects of non-rule-based clinical AI implementation. TRIAL REGISTRATION: PROSPERO (International Prospective Register of Systematic Reviews) CRD42021256005; https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=256005. INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): RR2-10.2196/33145.

From Data to Deployment: The Collaborative Community on Ophthalmic Imaging Roadmap for Artificial Intelligence in Age-Related Macular Degeneration.

OBJECTIVE: Health care systems worldwide are challenged to provide adequate care for the 200 million individuals with age-related macular degeneration (AMD). Artificial intelligence (AI) has the potential to make a significant, positive impact on the diagnosis and management of patients with AMD; however, the development of effective AI devices for clinical care faces numerous considerations and challenges, a fact evidenced by a current absence of Food and Drug Administration (FDA)-approved AI devices for AMD. PURPOSE: To delineate the state of AI for AMD, including current data, standards, achievements, and challenges. METHODS: Members of the Collaborative Community on Ophthalmic Imaging Working Group for AI in AMD attended an inaugural meeting on September 7, 2020, to discuss the topic. Subsequently, they undertook a comprehensive review of the medical literature relevant to the topic. Members engaged in meetings and discussion through December 2021 to synthesize the information and arrive at a consensus. RESULTS: Existing infrastructure for robust AI development for AMD includes several large, labeled data sets of color fundus photography and OCT images; however, image data often do not contain the metadata necessary for the development of reliable, valid, and generalizable models. Data sharing for AMD model development is made difficult by restrictions on data privacy and security, although potential solutions are under investigation. Computing resources may be adequate for current applications, but knowledge of machine learning development may be scarce in many clinical ophthalmology settings. Despite these challenges, researchers have produced promising AI models for AMD for screening, diagnosis, prediction, and monitoring. Future goals include defining benchmarks to facilitate regulatory authorization and subsequent clinical setting generalization. CONCLUSIONS: Delivering an FDA-authorized, AI-based device for clinical care in AMD involves numerous considerations, including the identification of an appropriate clinical application; acquisition and development of a large, high-quality data set; development of the AI architecture; training and validation of the model; and functional interactions between the model output and clinical end user. The research efforts undertaken to date represent starting points for the medical devices that eventually will benefit providers, health care systems, and patients.

Angiotensin-converting enzyme inhibitors and risk of age-related macular degeneration in individuals with hypertension.

AIMS: Several observational studies have examined the potential protective effect of angiotensin-converting enzyme inhibitor (ACE-I) use on the risk of age-related macular degeneration (AMD) and have reported contradictory results owing to confounding and time-related biases. We aimed to assess the risk of AMD in a base cohort of patients aged 40 years and above with hypertension among new users of ACE-I compared to an active comparator cohort of new users of calcium channel blockers (CCB) using data obtained from IQVIA Medical Research Data, a primary care database in the UK. METHODS: In this study, 53 832 and 43 106 new users of ACE-I and CCB were included between 1995 and 2019, respectively. In an on-treatment analysis, patients were followed up from the time of index drug initiation to the date of AMD diagnosis, loss to follow-up, discontinuation or switch to the comparator drug. A comprehensive range of covariates were used to estimate propensity scores to weight and match new users of ACE-I and CCB. Standardized mortality ratio weighted Cox proportional hazards model was used to estimate hazard ratios of developing AMD. RESULTS: During a median follow-up of 2 years (interquartile range 1-5 years), the incidence rate of AMD was 2.4 (95% confidence interval 2.2-2.6) and 2.2 (2.0-2.4) per 1000 person-years among the weighted new users of ACE-I and CCB, respectively. There was no association of ACE-I use on the risk of AMD compared to CCB use in either the propensity score weighted or matched, on-treatment analysis (adjusted hazard ratio: 1.07 [95% confidence interval 0.90-1.27] and 0.87 [0.71-1.07], respectively). CONCLUSION: We found no evidence that the use of ACE-I is associated with risk of AMD in patients with hypertension.

Retinal asymmetry in multiple sclerosis.

The diagnosis of multiple sclerosis is based on a combination of clinical and paraclinical tests. The potential contribution of retinal optical coherence tomography (OCT) has been recognized. We tested the feasibility of OCT measures of retinal asymmetry as a diagnostic test for multiple sclerosis at the community level. In this community-based study of 72 120 subjects, we examined the diagnostic potential of the inter-eye difference of inner retinal OCT data for multiple sclerosis using the UK Biobank data collected at 22 sites between 2007 and 2010. OCT reporting and quality control guidelines were followed. The inter-eye percentage difference (IEPD) and inter-eye absolute difference (IEAD) were calculated for the macular retinal nerve fibre layer (RNFL), ganglion cell inner plexiform layer (GCIPL) complex and ganglion cell complex. Area under the receiver operating characteristic curve (AUROC) comparisons were followed by univariate and multivariable comparisons accounting for a large range of diseases and co-morbidities. Cut-off levels were optimized by ROC and the Youden index. The prevalence of multiple sclerosis was 0.0023 [95% confidence interval (CI) 0.00229-0.00231]. Overall the discriminatory power of diagnosing multiple sclerosis with the IEPD AUROC curve (0.71, 95% CI 0.67-0.76) and IEAD (0.71, 95% CI 0.67-0.75) for the macular GCIPL complex were significantly higher if compared to the macular ganglion cell complex IEPD AUROC curve (0.64, 95% CI 0.59-0.69, P = 0.0017); IEAD AUROC curve (0.63, 95% CI 0.58-0.68, P < 0.0001) and macular RNFL IEPD AUROC curve (0.59, 95% CI 0.54-0.63, P < 0.0001); IEAD AUROC curve (0.55, 95% CI 0.50-0.59, P < 0.0001). Screening sensitivity levels for the macular GCIPL complex IEPD (4% cut-off) were 51.7% and for the IEAD (4 µm cut-off) 43.5%. Specificity levels were 82.8% and 86.8%, respectively. The number of co-morbidities was important. There was a stepwise decrease of the AUROC curve from 0.72 in control subjects to 0.66 in more than nine co-morbidities or presence of neuromyelitis optica spectrum disease. In the multivariable analyses greater age, diabetes mellitus, other eye disease and a non-white ethnic background were relevant confounders. For most interactions, the effect sizes were large (partial ω2 > 0.14) with narrow confidence intervals. In conclusion, the OCT macular GCIPL complex IEPD and IEAD may be considered as supportive measurements for multiple sclerosis diagnostic criteria in a young patient without relevant co-morbidity. The metric does not allow separation of multiple sclerosis from neuromyelitis optica. Retinal OCT imaging is accurate, rapid, non-invasive, widely available and may therefore help to reduce need for invasive and more costly procedures. To be viable, higher sensitivity and specificity levels are needed.

Sequence of clinical and neurodegeneration events in Parkinson's disease progression.

Dementia is one of the most debilitating aspects of Parkinson's disease. There are no validated biomarkers that can track Parkinson's disease progression, nor accurately identify patients who will develop dementia and when. Understanding the sequence of observable changes in Parkinson's disease in people at elevated risk for developing dementia could provide an integrated biomarker for identifying and managing individuals who will develop Parkinson's dementia. We aimed to estimate the sequence of clinical and neurodegeneration events, and variability in this sequence, using data-driven statistical modelling in two separate Parkinson's cohorts, focusing on patients at elevated risk for dementia due to their age at symptom onset. We updated a novel version of an event-based model that has only recently been extended to cope naturally with clinical data, enabling its application in Parkinson's disease for the first time. The observational cohorts included healthy control subjects and patients with Parkinson's disease, of whom those diagnosed at age 65 or older were classified as having high risk of dementia. The model estimates that Parkinson's progression in patients at elevated risk for dementia starts with classic prodromal features of Parkinson's disease (olfaction, sleep), followed by early deficits in visual cognition and increased brain iron content, followed later by a less certain ordering of neurodegeneration in the substantia nigra and cortex, neuropsychological cognitive deficits, retinal thinning in dopamine layers, and further deficits in visual cognition. Importantly, we also characterize variation in the sequence. We found consistent, cross-validated results within cohorts, and agreement between cohorts on the subset of features available in both cohorts. Our sequencing results add powerful support to the increasing body of evidence suggesting that visual processing specifically is affected early in patients with Parkinson's disease at elevated risk of dementia. This opens a route to earlier and more precise detection, as well as a more detailed understanding of the pathological mechanisms underpinning Parkinson's dementia.

Building trust in real-world data: lessons from INSIGHT, the UK's health data research hub for eye health and oculomics.

PURPOSE OF REVIEW: In this review, we consider the challenges of creating a trusted resource for real-world data in ophthalmology, based on our experience of establishing INSIGHT, the UK's Health Data Research Hub for Eye Health and Oculomics. RECENT FINDINGS: The INSIGHT Health Data Research Hub maximizes the benefits and impact of historical, patient-level UK National Health Service (NHS) electronic health record data, including images, through making it research-ready including curation and anonymisation. It is built around a shared 'north star' of enabling research for patient benefit. INSIGHT has worked to establish patient and public trust in the concept and delivery of INSIGHT, with efficient and robust governance processes that support safe and secure access to data for researchers. By linking to systemic data, there is an opportunity for discovery of novel ophthalmic biomarkers of systemic diseases ('oculomics'). Datasets that provide a representation of the whole population are an important tool to address the increasingly recognized threat of health data poverty. SUMMARY: Enabling efficient, safe access to routinely collected clinical data is a substantial undertaking, especially when this includes imaging modalities, but provides an exceptional resource for research. Research and innovation built on inclusive real-world data is an important tool in ensuring that discoveries and technologies of the future may not only favour selected groups, but also work for all patients.

Evaluating an automated machine learning model that predicts visual acuity outcomes in patients with neovascular age-related macular degeneration.

PURPOSE: Neovascular age-related macular degeneration (nAMD) is a major global cause of blindness. Whilst anti-vascular endothelial growth factor (anti-VEGF) treatment is effective, response varies considerably between individuals. Thus, patients face substantial uncertainty regarding their future ability to perform daily tasks. In this study, we evaluate the performance of an automated machine learning (AutoML) model which predicts visual acuity (VA) outcomes in patients receiving treatment for nAMD, in comparison to a manually coded model built using the same dataset. Furthermore, we evaluate model performance across ethnic groups and analyse how the models reach their predictions. METHODS: Binary classification models were trained to predict whether patients' VA would be 'Above' or 'Below' a score of 70 one year after initiating treatment, measured using the Early Treatment Diabetic Retinopathy Study (ETDRS) chart. The AutoML model was built using the Google Cloud Platform, whilst the bespoke model was trained using an XGBoost framework. Models were compared and analysed using the What-if Tool (WIT), a novel model-agnostic interpretability tool. RESULTS: Our study included 1631 eyes from patients attending Moorfields Eye Hospital. The AutoML model (area under the curve [AUC], 0.849) achieved a highly similar performance to the XGBoost model (AUC, 0.847). Using the WIT, we found that the models over-predicted negative outcomes in Asian patients and performed worse in those with an ethnic category of Other. Baseline VA, age and ethnicity were the most important determinants of model predictions. Partial dependence plot analysis revealed a sigmoidal relationship between baseline VA and the probability of an outcome of 'Above'. CONCLUSION: We have described and validated an AutoML-WIT pipeline which enables clinicians with minimal coding skills to match the performance of a state-of-the-art algorithm and obtain explainable predictions.

Quantitative Analysis of OCT for Neovascular Age-Related Macular Degeneration Using Deep Learning.

PURPOSE: To apply a deep learning algorithm for automated, objective, and comprehensive quantification of OCT scans to a large real-world dataset of eyes with neovascular age-related macular degeneration (AMD) and make the raw segmentation output data openly available for further research. DESIGN: Retrospective analysis of OCT images from the Moorfields Eye Hospital AMD Database. PARTICIPANTS: A total of 2473 first-treated eyes and 493 second-treated eyes that commenced therapy for neovascular AMD between June 2012 and June 2017. METHODS: A deep learning algorithm was used to segment all baseline OCT scans. Volumes were calculated for segmented features such as neurosensory retina (NSR), drusen, intraretinal fluid (IRF), subretinal fluid (SRF), subretinal hyperreflective material (SHRM), retinal pigment epithelium (RPE), hyperreflective foci (HRF), fibrovascular pigment epithelium detachment (fvPED), and serous PED (sPED). Analyses included comparisons between first- and second-treated eyes by visual acuity (VA) and race/ethnicity and correlations between volumes. MAIN OUTCOME MEASURES: Volumes of segmented features (mm3) and central subfield thickness (CST) (µm). RESULTS: In first-treated eyes, the majority had both IRF and SRF (54.7%). First-treated eyes had greater volumes for all segmented tissues, with the exception of drusen, which was greater in second-treated eyes. In first-treated eyes, older age was associated with lower volumes for RPE, SRF, NSR, and sPED; in second-treated eyes, older age was associated with lower volumes of NSR, RPE, sPED, fvPED, and SRF. Eyes from Black individuals had higher SRF, RPE, and serous PED volumes compared with other ethnic groups. Greater volumes of the majority of features were associated with worse VA. CONCLUSIONS: We report the results of large-scale automated quantification of a novel range of baseline features in neovascular AMD. Major differences between first- and second-treated eyes, with increasing age, and between ethnicities are highlighted. In the coming years, enhanced, automated OCT segmentation may assist personalization of real-world care and the detection of novel structure-function correlations. These data will be made publicly available for replication and future investigation by the AMD research community.

Longitudinal Development of Peripapillary Hyper-Reflective Ovoid Masslike Structures Suggests a Novel Pathological Pathway in Multiple Sclerosis.

OBJECTIVE: Peripapillary hyper-reflective ovoid masslike structures (PHOMS) are a new spectral domain optical coherence tomography (OCT) finding. METHODS: This prospective, longitudinal study included patients (n = 212) with multiple sclerosis (MS; n = 418 eyes), 59 healthy controls (HCs; n = 117 eyes), and 267 non-MS disease controls (534 eyes). OCT and diffusion tensor imaging were used. RESULTS: There were no PHOMS in HC eyes (0/117, 0%). The prevalence of PHOMS was significantly higher in patients with MS (34/212, p = 0.001) and MS eyes (45/418, p = 0.0002) when compared to HCs (0/59, 0/117). The inter-rater agreement for PHOMS was 97.9% (kappa = 0.951). PHOMS were present in 16% of patients with relapsing-remitting, 16% of patients with progressive, and 12% of patients with secondary progressive disease course (2% of eyes). There was no relationship of PHOMS with age, disease duration, disease course, disability, or disease-modifying treatments. The fractional anisotropy of the optic radiations was lower in patients without PHOMS (0.814) when compared to patients with PHOMS (0.845, p = 0.03). The majority of PHOMS remained stable, but increase in size and de novo development of PHOMS were also observed. In non-MS disease controls, PHOMS were observed in intracranial hypertension (62%), optic disc drusen (47%), anomalous optic discs (44%), isolated optic neuritis (19%), and optic atrophy (12%). INTERPRETATION: These data suggest that PHOMS are a novel finding in MS pathology. Future research is needed to determine whether development of PHOMS in MS is due to intermittently raised intracranial pressure or an otherwise impaired "glymphatic" outflow from eye to brain. ANN NEUROL 2020;88:309-319.

Automated deep learning in ophthalmology: AI that can build AI.

PURPOSE OF REVIEW: The purpose of this review is to describe the current status of automated deep learning in healthcare and to explore and detail the development of these models using commercially available platforms. We highlight key studies demonstrating the effectiveness of this technique and discuss current challenges and future directions of automated deep learning. RECENT FINDINGS: There are several commercially available automated deep learning platforms. Although specific features differ between platforms, they utilise the common approach of supervised learning. Ophthalmology is an exemplar speciality in the area, with a number of recent proof-of-concept studies exploring classification of retinal fundus photographs, optical coherence tomography images and indocyanine green angiography images. Automated deep learning has also demonstrated impressive results in other specialities such as dermatology, radiology and histopathology. SUMMARY: Automated deep learning allows users without coding expertise to develop deep learning algorithms. It is rapidly establishing itself as a valuable tool for those with limited technical experience. Despite residual challenges, it offers considerable potential in the future of patient management, clinical research and medical education. VIDEO ABSTRACT: http://links.lww.com/COOP/A44.

Applications of interpretability in deep learning models for ophthalmology.

PURPOSE OF REVIEW: In this article, we introduce the concept of model interpretability, review its applications in deep learning models for clinical ophthalmology, and discuss its role in the integration of artificial intelligence in healthcare. RECENT FINDINGS: The advent of deep learning in medicine has introduced models with remarkable accuracy. However, the inherent complexity of these models undermines its users' ability to understand, debug and ultimately trust them in clinical practice. Novel methods are being increasingly explored to improve models' 'interpretability' and draw clearer associations between their outputs and features in the input dataset. In the field of ophthalmology, interpretability methods have enabled users to make informed adjustments, identify clinically relevant imaging patterns, and predict outcomes in deep learning models. SUMMARY: Interpretability methods support the transparency necessary to implement, operate and modify complex deep learning models. These benefits are becoming increasingly demonstrated in models for clinical ophthalmology. As quality standards for deep learning models used in healthcare continue to evolve, interpretability methods may prove influential in their path to regulatory approval and acceptance in clinical practice.