Validation of a deep-learning-based retinal biomarker (Reti-CVD) in the prediction of cardiovascular disease: data from UK Biobank

Abstract

Background Currently in the United Kingdom, cardiovascular disease (CVD) risk assessment is based on the QRISK3 score, in which 10% 10-year CVD risk indicates clinical intervention. However, this benchmark has limited efficacy in clinical practice and the need for a more simple, non-invasive risk stratification tool is necessary. Retinal photography is becoming increasingly acceptable as a non-invasive imaging tool for CVD. Previously, we developed a novel CVD risk stratification system based on retinal photographs predicting future CVD risk. This study aims to further validate our biomarker, Reti-CVD, (1) to detect risk group of ≥ 10% in 10-year CVD risk and (2) enhance risk assessment in individuals with QRISK3 of 7.5–10% (termed as borderline-QRISK3 group) using the UK Biobank. Methods Reti-CVD scores were calculated and stratified into three risk groups based on optimized cut-off values from the UK Biobank. We used Cox proportional-hazards models to evaluate the ability of Reti-CVD to predict CVD events in the general population. C-statistics was used to assess the prognostic value of adding Reti-CVD to QRISK3 in borderline-QRISK3 group and three vulnerable subgroups. Results Among 48,260 participants with no history of CVD, 6.3% had CVD events during the 11-year follow-up. Reti-CVD was associated with an increased risk of CVD (adjusted hazard ratio [HR] 1.41; 95% confidence interval [CI], 1.30–1.52) with a 13.1% (95% CI, 11.7–14.6%) 10-year CVD risk in Reti-CVD-high-risk group. The 10-year CVD risk of the borderline-QRISK3 group was greater than 10% in Reti-CVD-high-risk group (11.5% in non-statin cohort [n = 45,473], 11.5% in stage 1 hypertension cohort [n = 11,966], and 14.2% in middle-aged cohort [n = 38,941]). C statistics increased by 0.014 (0.010–0.017) in non-statin cohort, 0.013 (0.007–0.019) in stage 1 hypertension cohort, and 0.023 (0.018–0.029) in middle-aged cohort for CVD event prediction after adding Reti-CVD to QRISK3. Conclusions Reti-CVD has the potential to identify individuals with ≥ 10% 10-year CVD risk who are likely to benefit from earlier preventative CVD interventions. For borderline-QRISK3 individuals with 10-year CVD risk between 7.5 and 10%, Reti-CVD could be used as a risk enhancer tool to help improve discernment accuracy, especially in adult groups that may be pre-disposed to CVD.

Prognostic potentials of AI in ophthalmology: systemic disease forecasting via retinal imaging

“…Adding Reti-CAC to the PCE enhances cardiovascular risk stratification, evidenced by a continuous net reclassification index of 0.261. Reti-CAC score Severance Main Hospital, Seoul, South Korea CMERC­HI (South Korea) SEED (Singapore) UK Biobank (UK) 527 (CMERC-HI) 33 (6.3%) of 527 had CVD events during the 5-year follow-up 8,551 (SEED) 310 (3.6%) of 8551 had CVD events 47,679 (UK Biobank) 337 (0.7%) of 47679 had CVD events Age Gender Hypertension Dyslipidaemia Diabetes Smoking status BMI Fasting glucose level Retrospective cohort study Tseng et al [ 17 ] May 2021 Reti-CVD, derived from retinal photographs, indicates a higher CVD risk, and serves as an effective tool for identifying individuals with a ≥ 10% 10-year CVD risk, including those in the borderline-QRISK3 category. It accurately identifies high-risk individuals, with a 10-year risk of 13.1% (CI: 11.7%–14.6%), enhancing early intervention and refining risk assessments, especially for those with a 10-year risk of 7.5%–10%.…”

Prediction of Cardiovascular Markers and Diseases Using Retinal Fundus Images and Deep Learning: A Systematic Scoping Review

“…Rim et al added the score to PCE-ASCVD to predict SCORE CVD events and Tseng et al added the score to QRISK3 to predict 10-year CVD respectively. 28,29 They both reported small improvements in predictive performance after adding retinal image-derived scores to the models. In another study, Reti-CVD was used to identify people at intermediate and high risk of CVD according to established clinical risk scores (PCE-ASCVD, QRISK3 and Framingham Risk Score), and reported that it could effectively identify these groups.…”

“…We found a series of studies on prediction of CAC from one research group, where they developed and validated a retinal fundus image-based score referred to as Reti-CVD or RetiCAC, and its utility was compared to existing CVD risk scores. [28][29][30] The score was proposed as a non-invasive and cost-efficient alternative to the computer tomography-derived CAC score. Multiple validation studies investigated the incremental predictive value of RetiCAC on top of established clinical risk prediction models.…”

“…The most common approach to circumvent this limitation was a two-step approach combining deep learning with classical survival analysis methods. 20,[28][29][30] In step one, the deep learning model takes one or a pair of retinal fundus images as input and outputs an intermediate score or abstract feature representation. In step two, the score is used, often in combination with clinical risk factors, as input for a Cox or Poisson regression model, that then outputs predicted probabilities.…”

“…Cross-sectional studies systolic and diastolic blood pressure 20,32,34,42 left ventricular characteristics 43 brachial-ankle pulse-wave velocity 44 coronary artery calcium score [28][29][30][45][46][47] carotid intima-media thickness 33 hypertension 48,49 peripheral arterial disease 50 coronary heart disease 10,51 cerebrovascular disease 10 stroke 19,[52][53][54] diagnosis of CVD 55 Follow-up studies major adverse cardiovascular events 32 peripheral arterial disease 20 coronary heart disease 20,29 stroke 20,21,29,30 myocardial infarction 20,21,30,43 major heart failure 21,30 CVD mortality [28][29][30]33…”