Author: Orlaith Mc Grath (United Kingdom)
Co-authors: Laura Steeples, Sasa Pockar, Tariq Aslam
Purpose
To develop an automated method for quantifying retinal ischaemia in Optical Coherence Tomography Angiography (OCTA), with particular application for use in the assessment of uveitic retinal ischaemia.
Setting/Venue
Manchester Royal Eye Hospital, Manchester, United Kingdom. Patients with uveitis and retinal ischaemia on widefield OCTA, determined by clinical experts, were included.
Methods
This was a feasibility study involving a retrospective, observational case-control series comprised of 10 patients (10 eyes), with a diagnosis of intermediate, posterior, or panuveitis and 10 patients (10 eyes) without intraocular inflammation or vitreoretinal disease. 10x10mm OCTA scans of the superficial capillary plexus were captured and denoised using inbuilt software on the Canon OCTA-1 machine. The images were anonymised and exported in binary format using proprietary Canon software. The exported images were analysed using a custom algorithm written by the last author in MATLAB 2021 (Mathworks, Natick, MA), which is fully automated and avoids the need for manual segmentation. Major blood vessels were removed using morphological techniques to purely segment the capillary network. Non-perfused areas were identified when they matched predefined thresholds of size, that were defined iteratively through experimentation. The identified ischaemic areas were represented as images for each patient’s OCTA with ischaemic areas filled in and output as a percentage of the ratio of pixels associated with non-perfused capillaries to total pixels. Ischemic areas identified on the images were compared to those identified by an Ophthalmologist. A paired t-Test was carried out using IBM SPSS Statistics 25 to compare percentage ischaemia in ischaemic vs normal eyes.
Results
The algorithm was able to identify ischaemic areas in all 10 patients who had been previously determined to show clinical ischaemia. Mean percentage of ischaemic areas in uveitic eyes was 3.9% vs the mean percentage in normal eyes of 0.03%. A paired t-test showed a significance level of 0.014 (2-tailed).
Conlusions
We describe a rapid automated method for objectively quantifying retinal ischaemia on OCT-A, developed and demonstrated on a discrete small number of patients. Our next step is to validate this novel algorithm on a larger, independent data set to truly demonstrate its validity and to investigate the role of this analysis in specific uveitis diseases involving the retinal circulation. Automated OCTA indices may provide objective markers of retinal ischaemia, which may be used in the evaluation and monitoring of ocular inflammatory disease with retinal ischaemia.
Financial Disclosure
Research Funding provided by Canon Medical Systems, Europe
Comments
-
