Detecting Retinal Hypoperfusion using Hyperspectral Imaging and Deep Learning

We will explore the application of novel hyperspectral imaging technology for the early detection of sight-threatening diseases. By capturing a rapid series of retinal images with different colours (wavelengths) of light and stacking these into an image cube, we create a rich source of data that can be interrogated with artificial intelligence methods to identify new disease markers.
Our preliminary studies have shown that this approach can detect areas of the retina with impaired blood supply (perfusion) with accuracy comparable to angiography, an invasive test that requires the intravenous injection of fluorescent dye. This advance has the potential to save sight, as impaired retinal perfusion is a key driver avoidable blindness from diabetic retinopathy and retinal vascular occlusion. Moreover, this approach may help to advance our understanding of the role that impaired retinal perfusion has in the pathogenesis of other common blinding diseases such as age-related macular degeneration and normal tension glaucoma