Defining Retinal and Choroidal Structures Using Hyperspectral Imaging

This investigator-initiated imaging study aims to evaluate the utility of hyperspectral retinal imaging (HSI) for the identification and characterisation of retinal and choroidal structures in both normal and diseased eyes.

Hyperspectral imaging is a non-invasive retinal imaging modality that captures sequential images across multiple wavelengths of light (typically \>25 spectral bands and up to approximately 90 wavelengths). This produces a three-dimensional dataset ("hypercube") containing spatial and spectral information for each pixel, enabling analysis of tissue-specific spectral reflectance properties. The technique differs from conventional retinal photography, which typically uses three colour channels (red, green, and blue), by providing significantly enhanced spectral resolution.

The study population will include approximately 1000 participants recruited from multiple ophthalmic clinics in Melbourne, Australia. Participants will be recruited across a range of normal ocular health and retinal disease states, including but not limited to diabetic retinopathy, glaucoma, and age-related macular degeneration.

Imaging will be performed at the Centre for Eye Research Australia (CERA) using two hyperspectral imaging systems: the Optina Diagnostics Metabolic Hyperspectral Retinal Camera and a prototype hyperspectral camera developed at CERA. Both systems acquire rapid sequential retinal images across visible to near-infrared wavelengths. The Optina device provides a field of view of approximately 26 degrees and uses a tunable supercontinuum light source, while the CERA prototype uses LED-based illumination with optical filtering and provides a field of view of approximately 35 degrees.

Participants may undergo pharmacological pupil dilation prior to imaging to improve image quality. Each study visit will last approximately 60 minutes, including dilation and imaging procedures.

Image data will undergo registration and processing to correct for eye movement and system response, enabling extraction of pixel-level spectral signatures. Computational analysis methods will be developed to identify and quantify retinal and choroidal features and to explore associations between spectral signatures and structural or functional measures of disease.

The primary objectives are to optimise hyperspectral imaging acquisition protocols for retinal and optic nerve structures, to determine whether spectral signatures associated with retinal pathology can be detected using hyperspectral imaging, and to assess whether these spectral changes correlate with clinical measures of disease severity.