Abstract
Purpose: :
To characterize hallmark diabetic retinopathy (DR) lesions utilizing an adaptive optics scanning laser ophthalmoscope (AOSLO) and to compare AOSLO findings with those from fundus photography and spectral domain optical coherence tomography (SDOCT).
Methods: :
Retinal imaging was performed in eyes of study subjects with DR using an AOSLO (Boston Micromachines) with 2μ resolution capable of fully compensating ocular optical aberrations in >90% of people. Each AOSLO image covered a 1°x1.2° area with 50 recorded frames. Image dewarping & averaging was performed using a customized Matlab platform (Burns lab, IN). Confocal AOSLO images were focused anteriorly at the retinal vessel level and posteriorly at the photoreceptor level. ETDRS 7 standard field digital stereo color fundus photography (ETDRS photos), infrared Spectralis macular imaging (IR imaging), and SDOCT (Spectralis, Heidelberg) were also obtained. Specific areas of diabetic pathology were compared between ETDRS photos, IR imaging, SDOCT and AOSLO videos and averaged images.
Results: :
Images were obtained from 15 eyes of 15 subjects. Three pathological lesions of DR have been evaluated to date: microaneurysms (MAs), hard exudates (HEs) and diabetic macular edema (DME). Although MAs could often not be clearly distinguished from small dot hemorrhages on ETDRS photos and SDOCT, some MAs could clearly be identified with AOSLO by virtue of observing RBC flow within feeding vessels as well as observing RBC flow within the MAs themselves. Localized dark patches beneath the MAs at the level of photoreceptors appeared to be a shadowing effect. Using AOSLO, HEs appeared as highly reflective confluent patches with a granular texture and darkened edges. When HE position within the retina was correlated using cross sectional SDOCT scans, HEs located in the inner retinal layers created shadows in the AOSLO image at the underlying photoreceptor layer whereas HEs in the outer retinal layers appeared hyperreflective regardless of image focal plane. In areas of DME, images focused throughout the retinal thickness had dark blurry patches. These patches varied in density, tending to have darker edges that might represent cyst walls as seen on SDOCT. Within areas of DME, photoreceptors could be imaged, but cell outlines were often blurred. Additional DR lesion types are being studied.
Conclusions: :
The AOSLO allows highly detailed in vivo imaging of DR lesions at the cellular level providing new morphological characteristics of these hallmark pathologies. This capability may eventually provide better means to identify sentinel lesions and associated disruption of key retinal cells such as photoreceptors.
Keywords: imaging/image analysis: clinical • diabetic retinopathy • imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound)