Assessing and monitoring abnormal retinal blood circulation for early detection of pre-structural damage and enhanced diabetic retinopathy staging using non-invasive, high-resolution biophotonic imaging technology - Eye

Barriers to earlier DR detection due to significant unmet needs in retinal oximetry Ophthalmologists use direct and indirect ophthalmoscopy to identify retinal structural damage and guide treatment if DR is present. Subtle vascular SO₂ changes are detectable only histologically. Fluorescein angiography reveals vascular abnormalities like microaneurysm leakage but is more effective in advanced DR [34]. OCT and Doppler OCT assess retinal structure and perfusion [35, 36], though superficial vascular SO₂ quantification with OCT remains unproven [37, 38]. While OCT is primarily structural, visible-light OCT (vis-OCT) enhances resolution and enables SO₂ measurement via haemoglobin absorption [37, 39]. Early vis-OCT studies in animals confirmed oxygen quantification under various conditions [37, 40], and Yi et al. applied vis-OCT in humans in 2015, detecting arterial-venous SO₂ differences near the optic nerve head [41, 42]. By 2020, vis-OCT measured microvascular oximetry in capillaries of rats and humans [43, 44]. Clinical adoption still faces hurdles, including signal-to-noise improvement, laser safety, and human optimisation [45,46,47]. Additionally, two advanced technologies have emerged recently. First, OCT angiography (OCTA) is a non-invasive imaging technique that visualises capillaries in all retinal layers by mapping blood cell movement based on changes in OCT signal intensity [48]. Second, IDx-DR, an AI-based system recently approved by the US FDA, enables early autonomous diagnosis of diabetic retinopathy (DR) using images from a fundus camera [49]. However, both devices primarily provide structural data and do not address the lack of functional data, such as real-time SO₂ measurements, which can detect changes before structural damage occurs [50, 51]. Nonetheless, the functional changes preceding the structural abnormalities are not detected by these methods, therefore limiting their benefit in early intervention. As discussed, early-stage treatment is generally effective in preserving vision and preventing severe complications [52]. Early treatment is necessary because functional alterations in the retina, such as variations in oxygen saturation, usually occur prior to structural lesions being seen [7]. Using retinal oximetry, a non-invasive technique, is a promising way of identifying these initial functional changes. Further, the detection of clinical DR is not typically performed using objective measures, as it is based on the subjective assessment of the clinician. Imaging oximetry has been applied in various contexts [2,3,4]. Hickam introduced the first fundus-based retinal oximetry using the Beer-Lambert Law to estimate SO₂ (~59% ±11%) from optical density at two wavelengths, but the method required constant calibration due to individual variability [53, 54]. Pittman advanced the approach using three wavelengths to minimise scattering effects [55], and Delori et al. measured haemoglobin extinction coefficients to estimate venous SO₂ (~45% ±7%) [56]. Retinal oximetry has since been used to study SO₂ across demographics and diseases [57,58,59,60,61,62,63,64,65,66,67,68,69,70,71], but conventional models produce wide error margins [72,73,74,75], limiting their ability to detect subtle (~2.5%) changes in retinal SO₂. In Europe, Oxymap uses spectroscopic imaging to measure retinal oxygen saturation [59, 76, 77]. Widely utilised in clinical research, it has been employed to study oximetry changes in various ocular diseases, including DR [12]. The device produces fundus images with colour-coded vascular segments, indicating oxygen levels in arteries and veins. Oxymap has been used in research on eye diseases such as retinitis pigmentosa, glaucoma, DR, and age-related macular degeneration (AMD), as well as systemic conditions like multiple sclerosis (MS), CNS sarcoidosis, and Alzheimer's disease. These studies highlight retinal oxygen saturation as a biomarker for both ocular and systemic conditions. Notably, one study found significantly higher oxygen saturation in DR patients compared to healthy individuals [7]. In normative studies, Oxymap demonstrated consistent results, with a standard deviation of 1.0% in arterioles and 1.4% in venules [78]. However, without additional data and confirmatory testing, the accuracy remains uncertain, as the measurements focus on vessels and not areas like the macula or optic disc, where early disease often manifests. Racial variability in oxygen saturation has been observed in studies, potentially due to methodological differences or the influence of fundus pigmentation [79,80,81]. Zilia in Canada uses the images from Oxymap in their AI-based technology to autonomously detect the early stage of NPDR more accurately [82]. Zilia also contributed to the finding that oxygen saturation on the optic nerve is significantly lower in glaucoma patients compared to normal subjects. Additionally, Zilia's ocular oximetry measurements showed a correlation between SO and retinal nerve fibre layer (RNFL) thickness, demnstrating some potential in monitoring glaucoma progression because RNFL thinning is an early indicator of optic nerve damage [83]. Further, Zilia and colleagues also studied variations in retinal oxygen saturation, particularly focusing on understanding these variations within a diverse, healthy population to establish normative data for future clinical applications [80]. Similarly, Imedos is also used for assessing retinal oxygen saturation [12]. While these tools are valuable for staging the disease based on structural changes, they have limitations in detecting DR prior to structural damages (pre-DR). Such a limitation underscores a significant gap between pre-DR and the intervention of DR.