Purpose:
Hyperspectral imaging offers a unique non-invasive technique to study retinal chromophores, such as the functional haemoglobin derivatives, relevant to assessing the metabolic status of the retina.This study aims to establish the ability of hyperspectral imaging to detect oximetric changes in the retinal vasculature of patients with retinovascular disease using novel image processing and spectral analysis techniques.
Methods:
A hyperspectral retinal imaging system consisting of a modified fundus camera, a tuneable filter and a low-noise CCD was used to capture sequential hyperspectral images of the human retina. A hyperspectral data cube with a spectral bandwidth of 500nm to 700nm obtained at wavelength steps of 2nm were acquired for each subject.Normal subjects (n=15) and patients with retinal vein occlusions (n=7), retinal artery occlusions (n=5) and proliferative diabetic retinopathy (n=3) were examined.Reflectance image processing techniques and a linear spectral unmixing algorithm were used to generate oximetric maps of the retinal vasculature.
Results:
Linear spectral unmixing produced consistent semi-quantitative oximetric maps of the retina in normal subjects (see figure left).In patients with retinovascular disease this technique detected consistent and clinically significant changes in vessel oxygenation principally in the venular circulation of the retina (see figure right).
Conclusions:
The analysis of hyperspectral retinal images is capable of accurately detecting oximetric changes in the retinal vasculature.The oximetric changes in the venular circulation in the diseased retina suggest either a reduced metabolic demand for oxygen in the retinal tissues or an arterio-venous shunting phenomenon within the retinal circulation.These techniques may be applied to the detection and monitoring of disease progression in patients with diabetic retinopathy and glaucoma.
Keywords: imaging/image analysis: clinical • oxygen • ischemia