AI-ENHANCED OCULOMICS REPORT

CFPs combined with superficial and deep enface OCTA images [32] or CFP-originated vasculometry [33, 34] have been used to generate AI algorithms for stroke prediction. The retinal age gap (difference between predicted biological age based on CFP and chronological age) has been used to predict a hazard ratio (HR) for stroke events, with the highest stratified population showing an HR of 2.37 [36]. White matter lesions (WMLs) from MRI, reflecting CeVD severity, can also be predicted from CFP-based algorithms [37]. Retinal imaging can thus be used to output CeVD brain imaging results, directly detect CeVD, or predict future stroke incidents.

Cheung et al. [38] used AI/DL algorithm to detect AD-dementia with an AUROC of 0.93, whereas detect amyloid β-positive AD with AUROC 0.73-0.85. Another study [39] developed a model by CFPs and clinical dementia rating global scores, showing narrower arteriolar diameter and wider venular diameter associated with higher risk of incident dementia. OCT images have been used to predict AD, Parkinson's disease (PD), and multiple sclerosis (MS) [40-42], and for long-term prediction of MS disability course [43]. OCTA images combined with OCT showed model efficacy in predicting mild cognitive impairment (MCI) with AUCs ranging from 0.693-0.960 [44].

Kidney failure is typically defined by estimated glomerular filtration rate (eGFR) levels. Sabanayagam et al. [48] used CFPs and a combined model trained with risk factors to predict CKD, achieving AUCs of 0.911 and 0.938. Kang et al. [49] predicted early renal function impairment with an AUC of 0.81. Liu et al. [55] used OCT images to predict DN, with an accuracy of 91.68%. The Reti-CKD score, derived from CFPs, and retinal age gap show promise as predictors of renal disease incidence, with stratified populations showing HRs as high as 9.36 compared to the lowest [56, 57].

Algorithms based on CFPs have shown good performance in predicting hypertension (AUCs of 0.766), hyperglycemia (AUCs of 0.880), and dyslipidemia (AUCs of 0.703) respectively [58]. Diabetes and diabetic peripheral neuropathy (DPN) can also be detected using AI-CFPs models [51, 59-61]. Hybrid models combining CFPs with traditional Chinese medicine characteristics have been explored to enhance prediction performance for diabetes [62]. Retinal age gap also correlates with metabolic syndrome and inflammation, indicating its promise as a variable [63].

Xiao et al. [64] used CFPs to generate a screening model for hepatobiliary diseases and an identification model for 6 specific hepatobiliary diseases (including liver cancer, liver cirrhosis, chronic viral hepatitis, non-alcoholic fatty liver disease, cholelithiasis, hepatic cyst). The AUROCs for screening and identification ranged from 0.62 to 0.84. Slit-lamp photos were also used to train models, achieving good performance.

Automated non-invasive anaemia screening based on retinal imaging is a breakthrough, especially for individuals with conditions such as diabetes, where anaemia can increase morbidity and mortality risks. Mitani et al. [65] used CFPs and metadata to predict Hb concentration and diagnose anemia, achieving an AUC of 0.88 for detecting anemia and 0.95 for moderate anaemia. Rim et al. [1] also used CFPs to predict Hb. UWF images [66] and OCT images [67, 68] have also been used for anaemia screening, achieving high accuracies.

AI-based retinal imaging predicts disorders across different systems through three pathways: estimating disease risk factors, predicting established examination results (like carotid ultrasound, cardiac CT, MRI), and directly estimating specific clinical diseases. Incorporating metadata or risk factors into hybrid algorithms improves predictive effectiveness. New variables like Reti-Age, Reti-CKD, and direct retinal parameters (caliber, geometry, fractals, tortuosity, artery-vein nicking) are quantitatively extracted by AI for correlation with clinical diseases. Longitudinal data further enables risk stratification and prediction of disease incidence and progression.

Challenges and Opportunities in AI-Oculomics