Deciphering the vasoreactivity signature in the human eye for personalised stratification and treatment response in cerebral small vessel disease

Background

Cerebral small vessel disease (SVD) is a major cause of stroke and dementia, but we lack accessible, individual-level markers of the microvascular dysfunction driving it. Markers of this process are essential to improving accessibility and equity of diagnosis, prognostication, and research into disease mechanisms across all groups in society, but cannot be delivered by traditional modalities like CT and MRI because of limited resolution and availability.

The retinal portion of the central nervous system vasculature may provide the high-resolution insights we need, since retinal vessel function is associated with structural and functional brain imaging in SVD (Blair, MacCormick et al., 2025). Dynamic Vessel Analyzer (DVA) recordings capture evolving caliber of retinal arterioles during rest and physiological challenge (Hanssen 2022; Petersen et al., 2022). However, current practice collapses these into a single value, discarding most of the information and potentially limiting our ability to understand disease at an individual level. Together, these multiple dynamic characteristics of central nervous system microvessels are likely to carry important information about neurovascular disease, but initiating the translational pathway towards scientific and clinical biomarkers requires a comprehensive method to quantify and compare retinal vessel reactivity between individuals and across time. 

Edinburgh is uniquely positioned for this project, since the only DVA in Scotland is currently creating longitudinal video data from two landmark cohorts of small vessel disease and cognitive impairment, MSS3 and LACI-Cog1, and this synergises with ongoing work on preclinical models of SVD in Edinburgh. Our retinal data therefore provide an essential translational middle-ground between traditional pre-clinical and clinical imaging.

Aims

Overall: Relate neurovascular physiology from the living human eye to personalised stratification in SVD, by combining signal processing, multivariate statistics, and multimodal clinical linkage.

Aim 1: Quantify a vasoreactivity signature from multiple waveform elements, and analyse it against natural history of disease, and response to randomised treatment (MSS3, LACI-Cog1).

Aim 2: Investigate spatiotemporal propagation in human retinal vessels with reference to preclinical data (from the van Veluw lab).

These will be approached through objectives:

a. Extract a feature vector from vessel calibre traces, including low frequency spontaneous vasomotion, rate of dilation in response to stimuli, time-to-peak, area under the curve, and post-stimulus constriction.

b. Create a parsimonious signature. Using e.g., shrinkage covariance estimation and regularised/dimension-reduction methods.

c. Track change over time against MRI SVD burden, and association with treatment, using multivariate inference and cross-validated predictive modelling.

d. Investigate evidence for wave propagation along vessels, with reference to pre-clinical model data, e.g., using continuous cross-correlation with sub-frame lag interpolation and hierarchical mixed-effects models of wave speed, direction, and conduction decay, to multiple vessel segments per eye. Propagation metrics will be evaluated as candidate markers of endothelial coupling and smooth muscle cell health and correlated with MRI perivascular-space burden and treatment. This could help us understand the role of vessel wave propagation in clearance of waste products from the central nervous system, which is crucial for tackling dementia.

These objectives are complementary but not strictly sequential. For example, (c) could be achieved to some extent without (a, b) using individual metrics. If necessary, an alternative objective of deep-learning vessel segmentation of raw video to recover denser, longer vessel traces could be targeted.

Training Outcomes

The student will gain skills across image analysis, applied mathematics and clinical imaging. This will include time–frequency signal processing (wavelets), multivariate and functional-data statistics, regularised/predictive modelling with proper cross-validation, and mixed-effects modelling of hierarchical repeated measures. They will develop domain expertise in retinal and cerebral microvascular physiology, MRI markers of SVD, and RCT biomarker design, supported by cross-disciplinary supervision spanning quantitative methods and clinical neurovascular science. Training in reproducible research and data management will underpin an open, scientifically useful analysis platform.

References: 

Blair GW, MacCormick IJC, McGrory S, et al (2025) Retinal vascular reactivity is associated with white matter hyperintensities and dysfunctional cerebrovascular reactivity in cerebral small vessel disease. J Cereb Blood Flow Metab 271678X251366079. https://doi.org/10.1177/0271678X251366079

Hanssen H, Streese L, Vilser W (2022) Retinal vessel diameters and function in cardiovascular risk and disease. Progress in Retinal and Eye Research 91:101095. https://doi.org/10.1016/j.preteyeres.2022.101095

Petersen L, Aalkjaer C, Bek T (2022) Vasomotion in Retinal Arterioles Is Modified by Exercise and Flicker Stimulation. Invest Ophthalmol Vis Sci 63:7. https://doi.org/10.1167/iovs.63.13.7