Association between deep learning measured retinal vessel calibre and incident myocardial infarction in a retrospective cohort from the UK Biobank.

BACKGROUND: Cardiovascular disease is a leading cause of global death. Prospective population-based studies have found that changes in retinal microvasculature are associated with the development of coronary artery disease. Recently, artificial intelligence deep learning (DL) algorithms have been developed for the fully automated assessment of retinal vessel calibres. METHODS: In this study, we validate the association between retinal vessel calibres measured by a DL system (Singapore I Vessel Assessment) and incident myocardial infarction (MI) and assess its incremental performance in discriminating patients with and without MI when added to risk prediction models, using a large UK Biobank cohort. RESULTS: Retinal arteriolar narrowing was significantly associated with incident MI in both the age, gender and fellow calibre-adjusted (HR=1.67 (95% CI: 1.19 to 2.36)) and multivariable models (HR=1.64 (95% CI: 1.16 to 2.32)) adjusted for age, gender and other cardiovascular risk factors such as blood pressure, diabetes mellitus (DM) and cholesterol status. The area under the receiver operating characteristic curve increased from 0.738 to 0.745 (p=0.018) in the age-gender-adjusted model and from 0.782 to 0.787 (p=0.010) in the multivariable model. The continuous net reclassification improvements (NRIs) were significant in the age and gender-adjusted (NRI=21.56 (95% CI: 3.33 to 33.42)) and the multivariable models (NRI=18.35 (95% CI: 6.27 to 32.61)). In the subgroup analysis, similar associations between retinal arteriolar narrowing and incident MI were observed, particularly for men (HR=1.62 (95% CI: 1.07 to 2.46)), non-smokers (HR=1.65 (95% CI: 1.13 to 2.42)), patients without DM (HR=1.73 (95% CI: 1.19 to 2.51)) and hypertensive patients (HR=1.95 (95% CI: 1.30 to 2.93)) in the multivariable models. CONCLUSION: Our results support DL-based retinal vessel measurements as markers of incident MI in a predominantly Caucasian population.

Comparative effectiveness of first-line antihypertensive drug classes on the maintenance of estimated glomerular filtration rate (eGFR) in real world primary care.

Renin-angiotensin system inhibitors (RASi), particularly angiotensin-converting enzyme inhibitors (ACEIs) and angiotensin II receptor blockers (ARBs), are commonly used in the treatment of hypertension and are recommended for kidney protection. Uncertainty remains about the effectiveness of RASi being used as first-line antihypertensive therapy on eGFR maintenance compared to its alternatives, especially for those with no or early-stage chronic kidney disease (CKD). We conducted a retrospective cohort study of 19,499 individuals (mean age 64.1, 43.5% males) from primary care in Singapore with 4.5 median follow-up years. The study cohort included newly diagnosed individuals with hypertension (whose eGFR was mainly in CKD stages G1-G2) and initiated on ACEIs, ARBs, beta-blockers (BBs), calcium channel blockers (CCBs) or diuretics (Ds) as first-line antihypertensive monotherapy. We compared the estimated glomerular filtration rate (eGFR) curve before/after the drug initiation over time of patients under different drug classes and analyzed the time to declining to a more advanced stage CKD. Inverse probability of treatment weighting (IPTW) was used to adjust for baseline confounding factors. Two key findings were observed. First, after initiating antihypertensive drugs, the eGFR almost maintained the same as the baseline in the first follow-up year, compared with dropping 3 mL/min/1.73 m2 per year before drug initiation. Second, ARBs were observed to be slightly inferior to ACEIs (HR = 1.14, 95% CI = (1.04, 1.23)) and other antihypertensive agents (HR = 1.10, 95% CI = (1.01, 1.20)) in delaying eGFR decline to a more advanced CKD stage in the study population. Our results showed that initiating antihypertensive agents can significantly maintain eGFR for those newly diagnosed patients with hypertension. However, RASi may not be superior to other antihypertensive agents in maintaining eGFR levels for non-CKD or early stages CKD patients.

Predictors of HbA1c treatment response to add-on medication following metformin monotherapy: a population-based cohort study.

Evidence on the influence of patient characteristics on HbA1c treatment response for add-on medications in patients with type 2 diabetes (T2D) is unclear. This study aims to investigate the predictors of HbA1c treatment response for three add-on medications (sulfonylureas (SU), dipeptidyl peptidase-4 (DPP-4) and sodium-glucose cotransporter-2 (SGLT-2) inhibitor) in metformin monotherapy treated patients with T2D. This retrospective cohort study was conducted using the electronic health record data from six primary care clinics in Singapore. A total of 9748 adult patients with T2D on metformin monotherapy receiving SU, DPP-4 or SGLT-2 add-on were 1:1 propensity score matched to patients receiving other add-on medications. Patient demographics, laboratory results, diabetes related complications, comedications, and treatment response at two endpoints (HbA1c reduction ≥ 1% at 6th month, HbA1c goal attainment < 7% at 12th month) were examined. Multiple logistic regression analyses were used to identify patient characteristics associated with the treatment responses. After matching, there were 1073, 517, and 290 paired cohorts of SU, DPP-4 and SGLT-2 respectively. Besides baseline HbA1c, patients with longer hypertension disease duration and higher cholesterol HDL were associated with better treatment response to SU medication add-on. Lower estimated glomerular filtration rate (eGFR), and angiotensin-II receptor medications were associated with better treatment response to DPP-4 add-on. Lower cholesterol HDL, higher creatinine serum, absence of renal complications and beta-blockers medications were associated with better treatment response to SGLT-2 add-on. The cholesterol HDL, creatinine serum, eGFR, hypertension disease duration, angiotensin-II receptors and beta-blockers class of medications can influence the HbA1c treatment response for SU, DPP-4 and SGLT-2 add-on medications. Knowing the patients' characteristics that influence treatment response can assist in guiding clinical decisions when selecting the appropriate add-on medication, ultimately helping to prevent the development of diabetes-related complications.

Influence of secondhand smoke exposure on the retinal vasculature of children in Hong Kong.

BACKGROUND: A recent prospective demonstrated that cardiovascular risk factors in early childhood were associated with later cardiovascular events. However, the impact of secondhand smoke (SHS) on children is unclear. The aims of this study is to determine the effects of SHS exposure on the retinal vasculature of children. METHODS: This is a population-based cross-sectional study of children aged 6 to 8 years. All participants received comprehensive ophthalmic examinations and retinal photography. Data on SHS exposure was derived from a validated questionnaire. A validated deep-learning system was used to automatically estimate retinal arteriolar and venular calibers from retinal photographs. Associations of quantitative retinal vessel caliber values with SHS exposure, number of smokers in the household, and total number of cigarettes smoked were determined by analyses of covariance (ANCOVA) after adjusting for potential confounders. Test of trend was determined by treating categorical risk factors as continuous ordinal variables. RESULTS: Here we show children exposed to SHS have wider retinal arteriolar (CRAE 152.1 µm vs. 151.3 µm, p < 0.001) and venular (CRVE 216.7 µm vs. 215.5 µm, p < 0.001) calibers compared to those in smoke-free homes, after adjustment for different factors. Wider arteriolar and venular calibers are also associated with increasing number of smokers in the family (p trend < 0.001) and more cigarettes smoked among family smokers (p trend<0.001). CONCLUSIONS: Exposure to SHS at home is associated with changes in retinal vasculature among children. This reinforces the adverse effect of secondhand smoking around children though further research incorporating comprehensive assessment of potential confounders is necessary.

Exposure to secondhand smoke can be harmful, particularly for our heart and lung health as adults. However, the impact of secondhand smoke on children is less clear. Here, we looked at the effects of secondhand smoke exposure on vessels within children's eyes. The health of these vessels is a potential indicator of overall eye health and is also associated with cardiovascular disease. Pictures were taken of children's eyes and analyzed using a computer program. We looked at the association between vessel measurements in the eye and how much secondhand smoke the children are exposed to. We observed differences in the vessels in children exposed to secondhand smoke, compared to those from smoke-free homes. These findings indicate that secondhand smoke may affect the health of children's eyes and highlight the need to promote smoke-free home environments.

Deep learning algorithms to detect diabetic kidney disease from retinal photographs in multiethnic populations with diabetes.

OBJECTIVE: To develop a deep learning algorithm (DLA) to detect diabetic kideny disease (DKD) from retinal photographs of patients with diabetes, and evaluate performance in multiethnic populations. MATERIALS AND METHODS: We trained 3 models: (1) image-only; (2) risk factor (RF)-only multivariable logistic regression (LR) model adjusted for age, sex, ethnicity, diabetes duration, HbA1c, systolic blood pressure; (3) hybrid multivariable LR model combining RF data and standardized z-scores from image-only model. Data from Singapore Integrated Diabetic Retinopathy Program (SiDRP) were used to develop (6066 participants with diabetes, primary-care-based) and internally validate (5-fold cross-validation) the models. External testing on 2 independent datasets: (1) Singapore Epidemiology of Eye Diseases (SEED) study (1885 participants with diabetes, population-based); (2) Singapore Macroangiopathy and Microvascular Reactivity in Type 2 Diabetes (SMART2D) (439 participants with diabetes, cross-sectional) in Singapore. Supplementary external testing on 2 Caucasian cohorts: (3) Australian Eye and Heart Study (AHES) (460 participants with diabetes, cross-sectional) and (4) Northern Ireland Cohort for the Longitudinal Study of Ageing (NICOLA) (265 participants with diabetes, cross-sectional). RESULTS: In SiDRP validation, area under the curve (AUC) was 0.826(95% CI 0.818-0.833) for image-only, 0.847(0.840-0.854) for RF-only, and 0.866(0.859-0.872) for hybrid. Estimates with SEED were 0.764(0.743-0.785) for image-only, 0.802(0.783-0.822) for RF-only, and 0.828(0.810-0.846) for hybrid. In SMART2D, AUC was 0.726(0.686-0.765) for image-only, 0.701(0.660-0.741) in RF-only, 0.761(0.724-0.797) for hybrid. DISCUSSION AND CONCLUSION: There is potential for DLA using retinal images as a screening adjunct for DKD among individuals with diabetes. This can value-add to existing DLA systems which diagnose diabetic retinopathy from retinal images, facilitating primary screening for DKD.

Risk of Normal Tension Glaucoma Progression From Automated Baseline Retinal-Vessel Caliber Analysis: A Prospective Cohort Study.

PURPOSE: To determine the relationship between baseline retinal-vessel calibers computed by a deep-learning system and the risk of normal tension glaucoma (NTG) progression. DESIGN: Prospective cohort study. METHODS: Three hundred and ninety eyes from 197 patients with NTG were followed up for at least 24 months. Retinal-vessel calibers (central retinal arteriolar equivalent [CRAE] and central retinal venular equivalent [CRVE]) were computed from fundus photographs at baseline using a previously validated deep-learning system. Retinal nerve fiber layer (RNFL) thickness and visual field (VF) were evaluated semiannually. The Cox proportional-hazards model was used to evaluate the relationship of baseline retinal-vessel calibers to the risk of glaucoma progression. RESULTS: Over a mean follow-up period of 34.36 ± 5.88 months, 69 NTG eyes (17.69%) developed progressive RNFL thinning and 22 eyes (5.64%) developed VF deterioration. In the multivariable Cox regression analysis adjusting for age, gender, intraocular pressure, mean ocular perfusion pressure, systolic blood pressure, axial length, standard automated perimetry mean deviation, and RNFL thickness, narrower baseline CRAE (hazard ratio per SD decrease [95% confidence interval], 1.36 [1.01-1.82]) and CRVE (1.35 [1.01-1.80]) were associated with progressive RNFL thinning and narrower baseline CRAE (1.98 [1.17-3.35]) was associated with VF deterioration. CONCLUSION: In this study, each SD decrease in the baseline CRAE or CRVE was associated with a more than 30% increase in the risk of progressive RNFL thinning and a more than 90% increase in the risk of VF deterioration during the follow-up period. Baseline attenuation of retinal vasculature in NTG eyes was associated with subsequent glaucoma progression. High-throughput deep-learning-based retinal vasculature analysis demonstrated its clinical utility for NTG risk assessment.

Using similar patients to predict complication in patients with diabetes, hypertension, and lipid disorder: a domain knowledge-infused convolutional neural network approach.

OBJECTIVE: This study aims to develop a convolutional neural network-based learning framework called domain knowledge-infused convolutional neural network (DK-CNN) for retrieving clinically similar patient and to personalize the prediction of macrovascular complication using the retrieved patients. MATERIALS AND METHODS: We use the electronic health records of 169 434 patients with diabetes, hypertension, and/or lipid disorder. Patients are partitioned into 7 subcohorts based on their comorbidities. DK-CNN integrates both domain knowledge and disease trajectory of patients over multiple visits to retrieve similar patients. We use normalized discounted cumulative gain (nDCG) and macrovascular complication prediction performance to evaluate the effectiveness of DK-CNN compared to state-of-the-art models. Ablation studies are conducted to compare DK-CNN with reduced models that do not use domain knowledge as well as models that do not consider short-term, medium-term, and long-term trajectory over multiple visits. RESULTS: Key findings from this study are: (1) DK-CNN is able to retrieve clinically similar patients and achieves the highest nDCG values in all 7 subcohorts; (2) DK-CNN outperforms other state-of-the-art approaches in terms of complication prediction performance in all 7 subcohorts; and (3) the ablation studies show that the full model achieves the highest nDCG compared with other 2 reduced models. DISCUSSION AND CONCLUSIONS: DK-CNN is a deep learning-based approach which incorporates domain knowledge and patient trajectory data to retrieve clinically similar patients. It can be used to assist physicians who may refer to the outcomes and past treatments of similar patients as a guide for choosing an effective treatment for patients.

Diabetes medication recommendation system using patient similarity analytics.

Type-2 diabetes mellitus (T2DM) is a medical condition in which oral medications avail to patients to curb their hyperglycaemia after failed dietary therapy. However, individual responses to the prescribed pharmacotherapy may differ due to their clinical profiles, comorbidities, lifestyles and medical adherence. One approach is to identify similar patients within the same community to predict their likely response to the prescribed diabetes medications. This study aims to present an evidence-based diabetes medication recommendation system (DMRS) underpinned by patient similarity analytics. The DMRS was developed using 10-year electronic health records of 54,933 adult patients with T2DM from six primary care clinics in Singapore. Multiple clinical variables including patient demographics, comorbidities, laboratory test results, existing medications, and trajectory patterns of haemoglobin A1c (HbA1c) were used to identify similar patients. The DMRS was evaluated on four groups of patients with comorbidities such as hyperlipidaemia (HLD) and hypertension (HTN). Recommendations were assessed using hit ratio which represents the percentage of patients with at least one recommended sets of medication matches exactly the diabetes prescriptions in both the type and dosage. Recall, precision, and mean reciprocal ranking of the recommendation against the diabetes prescriptions in the EHR records were also computed. Evaluation against the EHR prescriptions revealed that the DMRS recommendations can achieve hit ratio of 81% for diabetes patients with no comorbidity, 84% for those with HLD, 78% for those with HTN, and 75% for those with both HLD and HTN. By considering patients' clinical profiles and their trajectory patterns of HbA1c, the DMRS can provide an individualized recommendation that resembles the actual prescribed medication and dosage. Such a system is useful as a shared decision-making tool to assist clinicians in selecting the appropriate medications for patients with T2DM.

Deep-learning retinal vessel calibre measurements and risk of cognitive decline and dementia.

Previous studies have explored the associations of retinal vessel calibre, measured from retinal photographs or fundus images using semi-automated computer programs, with cognitive impairment and dementia, supporting the concept that retinal blood vessels reflect microvascular changes in the brain. Recently, artificial intelligence deep-learning algorithms have been developed for the fully automated assessment of retinal vessel calibres. Therefore, we aimed to determine whether deep-learning-based retinal vessel calibre measurements are predictive of risk of cognitive decline and dementia. We conducted a prospective study recruiting participants from memory clinics at the National University Hospital and St. Luke's Hospital in Singapore; all participants had comprehensive clinical and neuropsychological examinations at baseline and annually for up to 5 years. Fully automated measurements of retinal arteriolar and venular calibres from retinal fundus images were estimated using a deep-learning system. Cox regression models were then used to assess the relationship between baseline retinal vessel calibre and the risk of cognitive decline and developing dementia, adjusting for age, gender, ethnicity, education, cerebrovascular disease status, hypertension, hyperlipidemia, diabetes, and smoking. A total of 491 participants were included in this study, of whom 254 developed cognitive decline over 5 years. In multivariable models, narrower retinal arteriolar calibre (hazard ratio per standard deviation decrease = 1.258, P = 0.008) and wider retinal venular calibre (hazard ratio per standard deviation increase = 1.204, P = 0.037) were associated with increased risk of cognitive decline. Among participants with cognitive impairment but no dementia at baseline (n = 212), 44 progressed to have incident dementia; narrower retinal arteriolar calibre was also associated with incident dementia (hazard ratio per standard deviation decrease = 1.624, P = 0.021). In summary, deep-learning-based measurement of retinal vessel calibre was associated with risk of cognitive decline and dementia.

Glycaemic control of Asian patients with type-2 diabetes mellitus on tiered up-titration of metformin monotherapy: A one-year real-world retrospective longitudinal study in primary care.

AIMS: To determine the glycaemic control and associated factors among patients with type-2 diabetes mellitus on tiered metformin monotherapy over one-year. METHODS: Adult Asian patients on metformin monotherapy with tiered dosage up-titration (low < 500 mg/day; medium 500-<1000 mg/day and high ≥ 1000 mg/day) are divided into four sub-cohorts based on their baseline HbA1c < 7%(C<7); 7%-<8%(C7-<8); 8%-<9%(C8-<9) and ≥ 9%(C≥9). The HbA1c absolute reduction, time to reach glycaemic control (HbA1c < 7%), and time from glycaemic control to failure (HbA1c ≥ 7%) after the dosage up-titration were the outcomes. RESULTS: Among 5503 eligible patients (mean age = 64.9 years, 45.6% males and 74.6% Chinese), the HbA1c absolute reduction after the up-titration at three months are 0%, 0.4%-0.6%, 0.8%-1.2% and 2.0%-2.1% for C<7, C7-<8, C8-<9 and C≥9 respectively. The median time (months) to attain glycaemic control for low, medium and high dosage up-titration were 4, 3, 3(C7-<8); 12, 7, 4(C8-<9); NA, 7, 7(C≥9). Within twelve months after the goal attainment, 36.2%(C<7), 48.8%(C7-<8), 52.7%(C8-<9) and 45.3%(C≥9) of patients had treatment failure. CONCLUSIONS: The results show that the baseline HbA1c and tiered metformin dosage up-titration are associated with disproportionate HbA1c reduction, time to glycaemic control and time from glycaemic control to failure.

The effect of oral diabetes medications on glycated haemoglobin (HbA1c) in Asians in primary care: a retrospective cohort real-world data study.

BACKGROUND: Clinical trials have demonstrated that initiating oral anti-diabetic drugs (OADs) significantly reduce glycated hemoglobin (HbA1c) levels. However, variability in lifestyle modifications and OAD adherence impact on their actual effect on glycemic control. Furthermore, evidence on dose adjustments and discontinuation of OAD on HbA1c is lacking. This study aims to use real-world data to determine the effect of OAD initiation, up-titration, down-titration, and discontinuation on HbA1c levels, among Asian patients managed in primary care. METHODS: A retrospective cohort study over a 5-year period, from Jan 2015 to Dec 2019 was conducted on a cohort of multi-ethnic adult Asian patients with clinical diagnosis of type 2 diabetes mellitus (T2DM) managed by a network of primary care clinics in Singapore. Nine OADs from five different classes (biguanides, sulphonyurea, dipeptidyl peptidase-4 [DPP-4] inhibitors, sodium-glucose cotransporter-2 [SGLT-2] inhibitors, and alpha-glucosidase inhibitors) were evaluated. Patients were grouped into "No OAD", "Non-titrators," and "Titrators" cohorts based on prescribing patterns. For the "Titrators" cohort, the various OAD titrations were identified. Subsequently, a descriptive analysis of HbA1c values before and after each titration was performed to compute a mean difference for each unique titration identified. RESULTS: Among the cohort of 57,910 patients, 43,338 of them had at least one OAD titration, with a total of 76,990 pairs of HbA1c values associated with an OAD titration. There were a total of 206 unique OAD titrations. Overall, initiation of OADs resulted in a reduction of HbA1c by 3 to 12 mmol/mol (0.3 to 1.1%), respectively. These results were slightly lower than those reported in clinical trials of 6 to 14 mmol/mol (0.5 to 1.25%). The change of HbA1c levels due to up-titration, down-titration, and discontinuation were -1 to -8 mmol/mol (-0.1 to -0.7%), +1 to 7 mmol/mol (+0.1 to +0.6%), and +2 to 11 mmol/mol (+0.2 to +1.0%), respectively. The HbA1c lowering effect of initiating newer OADs, namely DPP-4 inhibitors and SGLT-2 inhibitors was 8 to 11 mmol/mol (0.7 to 0.9%) and 7 to 11 mmol/mol (0.6 to 1.0%), respectively. CONCLUSION: The real-world data on Asians with T2DM in this study show that the magnitudes of OAD initiation and dose titration are marginally lower than the results from clinical trials. During shared decision-making in selecting treatment options, the results enable physicians to communicate realistic expectation of the effect of oral medications on the glycemic control of their patients in primary care.

Using Domain Knowledge and Data-Driven Insights for Patient Similarity Analytics.

Patient similarity analytics has emerged as an essential tool to identify cohorts of patients who have similar clinical characteristics to some specific patient of interest. In this study, we propose a patient similarity measure called D3K that incorporates domain knowledge and data-driven insights. Using the electronic health records (EHRs) of 169,434 patients with either diabetes, hypertension or dyslipidaemia (DHL), we construct patient feature vectors containing demographics, vital signs, laboratory test results, and prescribed medications. We discretize the variables of interest into various bins based on domain knowledge and make the patient similarity computation to be aligned with clinical guidelines. Key findings from this study are: (1) D3K outperforms baseline approaches in all seven sub-cohorts; (2) our domain knowledge-based binning strategy outperformed the traditional percentile-based binning in all seven sub-cohorts; (3) there is substantial agreement between D3K and physicians (κ = 0.746), indicating that D3K can be applied to facilitate shared decision making. This is the first study to use patient similarity analytics on a cardiometabolic syndrome-related dataset sourced from medical institutions in Singapore. We consider patient similarity among patient cohorts with the same medical conditions to develop localized models for personalized decision support to improve the outcomes of a target patient.

Patient similarity analytics for explainable clinical risk prediction.

BACKGROUND: Clinical risk prediction models (CRPMs) use patient characteristics to estimate the probability of having or developing a particular disease and/or outcome. While CRPMs are gaining in popularity, they have yet to be widely adopted in clinical practice. The lack of explainability and interpretability has limited their utility. Explainability is the extent of which a model's prediction process can be described. Interpretability is the degree to which a user can understand the predictions made by a model. METHODS: The study aimed to demonstrate utility of patient similarity analytics in developing an explainable and interpretable CRPM. Data was extracted from the electronic medical records of patients with type-2 diabetes mellitus, hypertension and dyslipidaemia in a Singapore public primary care clinic. We used modified K-nearest neighbour which incorporated expert input, to develop a patient similarity model on this real-world training dataset (n = 7,041) and validated it on a testing dataset (n = 3,018). The results were compared using logistic regression, random forest (RF) and support vector machine (SVM) models from the same dataset. The patient similarity model was then implemented in a prototype system to demonstrate the identification, explainability and interpretability of similar patients and the prediction process. RESULTS: The patient similarity model (AUROC = 0.718) was comparable to the logistic regression (AUROC = 0.695), RF (AUROC = 0.764) and SVM models (AUROC = 0.766). We packaged the patient similarity model in a prototype web application. A proof of concept demonstrated how the application provided both quantitative and qualitative information, in the form of patient narratives. This information was used to better inform and influence clinical decision-making, such as getting a patient to agree to start insulin therapy. CONCLUSIONS: Patient similarity analytics is a feasible approach to develop an explainable and interpretable CRPM. While the approach is generalizable, it can be used to develop locally relevant information, based on the database it searches. Ultimately, such an approach can generate a more informative CRPMs which can be deployed as part of clinical decision support tools to better facilitate shared decision-making in clinical practice.

LDL-cholesterol change and goal attainment following statin intensity titration among Asians in primary care: a retrospective cohort study.

BACKGROUND: Clinical trials have demonstrated that either initiating or up-titrating a statin dose substantially reduce Low-Density Lipoprotein-Cholesterol (LDL-C) levels. However, statin adherence in actual practice tends to be suboptimal, leading to diminished effectiveness. This study aims to use real-world data to determine the effect on LDL-C levels and LDL-C goal attainment rates, when selected statins are titrated in Asian patients. METHODS: A retrospective cohort study over a 5-year period, from April 2014 to March 2019 was conducted on a cohort of multi-ethnic adult Asian patients with clinical diagnosis of Dyslipidaemia in a primary care clinic in Singapore. The statins were classified into low-intensity (LI), moderate-intensity (MI) and high-intensity (HI) groups according to the 2018 American College of Cardiology and American Heart Association (ACC/AHA) Blood Cholesterol Guidelines. Patients were grouped into "No statin", "Non-titrators" and "Titrators" cohorts based on prescribing patterns. For the "Titrators" cohort, the mean percentage change in LDL-C and absolute change in LDL-C goal attainment rates were computed for each permutation of statin intensity titration. RESULTS: Among the cohort of 11,499 patients, with a total of 266,762 visits, there were 1962 pairs of LDL-C values associated with a statin titration. Initiation of LI, MI and HI statin resulted in a lowering of LDL-C by 21.6% (95%CI = 18.9-24.3%), 28.9% (95%CI = 25.0-32.7%) and 25.2% (95%CI = 12.8-37.7%) respectively. These were comparatively lower than results from clinical trials (30 to 63%). The change of LDL-C levels due to up-titration, down-titration, and discontinuation were - 12.4% to - 28.9%, + 13.2% to + 24.6%, and + 18.1% to + 32.1% respectively. The improvement in LDL-C goal attainment ranged from 26.5% to 47.1% when statin intensity was up-titrated. CONCLUSION: In this study based on real-world data of Asian patients in primary care, it was shown that although statin titration substantially affected LDL-C levels and LDL-C goal attainment rates, the magnitude was lower than results reported from clinical trials. These results should be taken into consideration and provide further insight to clinicians when making statin adjustment recommendations in order to achieve LDL-C targets in clinical practice, particularly for Asian populations.

A deep-learning system for the assessment of cardiovascular disease risk via the measurement of retinal-vessel calibre.

Retinal blood vessels provide information on the risk of cardiovascular disease (CVD). Here, we report the development and validation of deep-learning models for the automated measurement of retinal-vessel calibre in retinal photographs, using diverse multiethnic multicountry datasets that comprise more than 70,000 images. Retinal-vessel calibre measured by the models and by expert human graders showed high agreement, with overall intraclass correlation coefficients of between 0.82 and 0.95. The models performed comparably to or better than expert graders in associations between measurements of retinal-vessel calibre and CVD risk factors, including blood pressure, body-mass index, total cholesterol and glycated-haemoglobin levels. In retrospectively measured prospective datasets from a population-based study, baseline measurements performed by the deep-learning system were associated with incident CVD. Our findings motivate the development of clinically applicable explainable end-to-end deep-learning systems for the prediction of CVD on the basis of the features of retinal vessels in retinal photographs.

A deep learning algorithm to detect chronic kidney disease from retinal photographs in community-based populations.

BACKGROUND: Screening for chronic kidney disease is a challenge in community and primary care settings, even in high-income countries. We developed an artificial intelligence deep learning algorithm (DLA) to detect chronic kidney disease from retinal images, which could add to existing chronic kidney disease screening strategies. METHODS: We used data from three population-based, multiethnic, cross-sectional studies in Singapore and China. The Singapore Epidemiology of Eye Diseases study (SEED, patients aged ≥40 years) was used to develop (5188 patients) and validate (1297 patients) the DLA. External testing was done on two independent datasets: the Singapore Prospective Study Program (SP2, 3735 patients aged ≥25 years) and the Beijing Eye Study (BES, 1538 patients aged ≥40 years). Chronic kidney disease was defined as estimated glomerular filtration rate less than 60 mL/min per 1·73m2. Three models were trained: 1) image DLA; 2) risk factors (RF) including age, sex, ethnicity, diabetes, and hypertension; and 3) hybrid DLA combining image and RF. Model performances were evaluated using the area under the receiver operating characteristic curve (AUC). FINDINGS: In the SEED validation dataset, the AUC was 0·911 for image DLA (95% CI 0·886 -0·936), 0·916 for RF (0·891-0·941), and 0·938 for hybrid DLA (0·917-0·959). Corresponding estimates in the SP2 testing dataset were 0·733 for image DLA (95% CI 0·696-0·770), 0·829 for RF (0·797-0·861), and 0·810 for hybrid DLA (0·776-0·844); and in the BES testing dataset estimates were 0·835 for image DLA (0·767-0·903), 0·887 for RF (0·828-0·946), and 0·858 for hybrid DLA (0·794-0·922). AUC estimates were similar in subgroups of people with diabetes (image DLA 0·889 [95% CI 0·850-0·928], RF 0·899 [0·862-0·936], hybrid 0·925 [0·893-0·957]) and hypertension (image DLA 0·889 [95% CI 0·860-0·918], RF 0·889 [0·860-0·918], hybrid 0·918 [0·893-0·943]). INTERPRETATION: A retinal image DLA shows good performance for estimating chronic kidney disease, underlying the feasibility of using retinal photography as an adjunctive or opportunistic screening tool for chronic kidney disease in community populations. FUNDING: National Medical Research Council, Singapore.

Technical and imaging factors influencing performance of deep learning systems for diabetic retinopathy.

Deep learning (DL) has been shown to be effective in developing diabetic retinopathy (DR) algorithms, possibly tackling financial and manpower challenges hindering implementation of DR screening. However, our systematic review of the literature reveals few studies studied the impact of different factors on these DL algorithms, that are important for clinical deployment in real-world settings. Using 455,491 retinal images, we evaluated two technical and three image-related factors in detection of referable DR. For technical factors, the performances of four DL models (VGGNet, ResNet, DenseNet, Ensemble) and two computational frameworks (Caffe, TensorFlow) were evaluated while for image-related factors, we evaluated image compression levels (reducing image size, 350, 300, 250, 200, 150 KB), number of fields (7-field, 2-field, 1-field) and media clarity (pseudophakic vs phakic). In detection of referable DR, four DL models showed comparable diagnostic performance (AUC 0.936-0.944). To develop the VGGNet model, two computational frameworks had similar AUC (0.936). The DL performance dropped when image size decreased below 250 KB (AUC 0.936, 0.900, p < 0.001). The DL performance performed better when there were increased number of fields (dataset 1: 2-field vs 1-field-AUC 0.936 vs 0.908, p < 0.001; dataset 2: 7-field vs 2-field vs 1-field, AUC 0.949 vs 0.911 vs 0.895). DL performed better in the pseudophakic than phakic eyes (AUC 0.918 vs 0.833, p < 0.001). Various image-related factors play more significant roles than technical factors in determining the diagnostic performance, suggesting the importance of having robust training and testing datasets for DL training and deployment in the real-world settings.

Artificial Intelligence Screening for Diabetic Retinopathy: the Real-World Emerging Application.

PURPOSE OF REVIEW: This paper systematically reviews the recent progress in diabetic retinopathy screening. It provides an integrated overview of the current state of knowledge of emerging techniques using artificial intelligence integration in national screening programs around the world. Existing methodological approaches and research insights are evaluated. An understanding of existing gaps and future directions is created. RECENT FINDINGS: Over the past decades, artificial intelligence has emerged into the scientific consciousness with breakthroughs that are sparking increasing interest among computer science and medical communities. Specifically, machine learning and deep learning (a subtype of machine learning) applications of artificial intelligence are spreading into areas that previously were thought to be only the purview of humans, and a number of applications in ophthalmology field have been explored. Multiple studies all around the world have demonstrated that such systems can behave on par with clinical experts with robust diagnostic performance in diabetic retinopathy diagnosis. However, only few tools have been evaluated in clinical prospective studies. Given the rapid and impressive progress of artificial intelligence technologies, the implementation of deep learning systems into routinely practiced diabetic retinopathy screening could represent a cost-effective alternative to help reduce the incidence of preventable blindness around the world.

Deep learning in estimating prevalence and systemic risk factors for diabetic retinopathy: a multi-ethnic study.

In any community, the key to understanding the burden of a specific condition is to conduct an epidemiological study. The deep learning system (DLS) recently showed promising diagnostic performance for diabetic retinopathy (DR). This study aims to use DLS as the grading tool, instead of human assessors, to determine the prevalence and the systemic cardiovascular risk factors for DR on fundus photographs, in patients with diabetes. This is a multi-ethnic (5 races), multi-site (8 datasets from Singapore, USA, Hong Kong, China and Australia), cross-sectional study involving 18,912 patients (n = 93,293 images). We compared these results and the time taken for DR assessment by DLS versus 17 human assessors - 10 retinal specialists/ophthalmologists and 7 professional graders). The estimation of DR prevalence between DLS and human assessors is comparable for any DR, referable DR and vision-threatening DR (VTDR) (Human assessors: 15.9, 6.5% and 4.1%; DLS: 16.1%, 6.4%, 3.7%). Both assessment methods identified similar risk factors (with comparable AUCs), including younger age, longer diabetes duration, increased HbA1c and systolic blood pressure, for any DR, referable DR and VTDR (p > 0.05). The total time taken for DLS to evaluate DR from 93,293 fundus photographs was ~1 month compared to 2 years for human assessors. In conclusion, the prevalence and systemic risk factors for DR in multi-ethnic population could be determined accurately using a DLS, in significantly less time than human assessors. This study highlights the potential use of AI for future epidemiology or clinical trials for DR grading in the global communities.

Validation of a Natural Language Processing Algorithm for Detecting Infectious Disease Symptoms in Primary Care Electronic Medical Records in Singapore.

BACKGROUND: Free-text clinical records provide a source of information that complements traditional disease surveillance. To electronically harness these records, they need to be transformed into codified fields by natural language processing algorithms. OBJECTIVE: The aim of this study was to develop, train, and validate Clinical History Extractor for Syndromic Surveillance (CHESS), an natural language processing algorithm to extract clinical information from free-text primary care records. METHODS: CHESS is a keyword-based natural language processing algorithm to extract 48 signs and symptoms suggesting respiratory infections, gastrointestinal infections, constitutional, as well as other signs and symptoms potentially associated with infectious diseases. The algorithm also captured the assertion status (affirmed, negated, or suspected) and symptom duration. Electronic medical records from the National Healthcare Group Polyclinics, a major public sector primary care provider in Singapore, were randomly extracted and manually reviewed by 2 human reviewers, with a third reviewer as the adjudicator. The algorithm was evaluated based on 1680 notes against the human-coded result as the reference standard, with half of the data used for training and the other half for validation. RESULTS: The symptoms most commonly present within the 1680 clinical records at the episode level were those typically present in respiratory infections such as cough (744/7703, 9.66%), sore throat (591/7703, 7.67%), rhinorrhea (552/7703, 7.17%), and fever (928/7703, 12.04%). At the episode level, CHESS had an overall performance of 96.7% precision and 97.6% recall on the training dataset and 96.0% precision and 93.1% recall on the validation dataset. Symptoms suggesting respiratory and gastrointestinal infections were all detected with more than 90% precision and recall. CHESS correctly assigned the assertion status in 97.3%, 97.9%, and 89.8% of affirmed, negated, and suspected signs and symptoms, respectively (97.6% overall accuracy). Symptom episode duration was correctly identified in 81.2% of records with known duration status. CONCLUSIONS: We have developed an natural language processing algorithm dubbed CHESS that achieves good performance in extracting signs and symptoms from primary care free-text clinical records. In addition to the presence of symptoms, our algorithm can also accurately distinguish affirmed, negated, and suspected assertion statuses and extract symptom durations.