Abstract
Purpose :
Hyperspectral (HS) cameras image across many wavelengths to detect retinal tissue reflectance changes. An increased risk of retinal pathology with age is paralleled by cataract formation which affects the measured light spectrum and hampers the utility of HS imaging in disease diagnosis and monitoring. While the lens’ bulk transmittance properties are known, the spectral influence of cataract subtypes (nuclear sclerotic, cortical, posterior subcapsular) and their spatial influence on HS images are poorly understood. We performed a prospective clinical study to ascertain if the spectral influence of various types of cataracts is well-defined.
Methods :
45 participants (20 male, 25 female, mean age±SD, 68.9±8.8) booked for cataract surgery were recruited. Exclusion criteria were diabetic retinopathy, glaucoma, age-related macular degeneration, eye surgery <6 months prior, previous eye trauma. The surgery eye was imaged using a Metabolic Hyperspectral Retinal Camera (450-900 nm, 5 nm steps, 30° field-of-view, Optina Diagnostics, Montreal, Canada), prior to and 4-weeks after surgery. Cataracts were graded using the Lens Opacities Classification System III. Each participant's HS images were co-registered and pixels with anomalous variation in reflectance were removed. Non-negative matrix factorisation (NMF) was performed on the change in retinal reflectance to extract spectral features explaining a large percentage of the data variance.
Results :
Two NMF spectral features over 450-900 nm modelled up to 95% of the variance of the data. The first feature was strongly weighted <650 nm, and may be attributed to nuclear sclerosis. The second feature was moderately weighted across all wavelengths, and may be attributed to cortical cataract. Across all wavelengths, there was a statistically significant increase in retinal reflectance post-surgery by a factor of 1.4 (95% CI: 1.34-1.50, p<0.0001). The maximum change was observed at 450 nm, where retinal reflectance increased post-surgery by a factor of 2.9 (95% CI: 2.54-3.24, p<0.0001).
Conclusions :
The spectral effect of cataracts on HS images was well-defined, and the retinal reflectance could be modelled as a mixture of two spectral features, which may be attributed to nuclear sclerotic and cortical cataract. These findings will facilitate developing techniques to correct for cataract in HS images, and subsequently detect spectral changes due to retinal pathology.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.