Image-processing pipeline for diagnosing diabetic macular ischaemia in diabetic macular oedema

Diabetic Macular Ischaemia (DMI) is a critical cause of vision loss, frequently occurring alongside Diabetic Macular Oedema (DMO) in patients with severe diabetic retinopathy. Accurate diagnosis of DMI is essential for assessing the visual prognosis of any intervention. In the context of clinical research for developing novel treatments to improve macular capillary perfusion, traditional imaging techniques often struggle to identify ischaemic areas precisely. Optical Coherence Tomography Angiography (OCTA) offers a non-invasive method for visualising retinal vasculature, providing potential improvements in diagnosis. However, reliable quantification of ischemia remains challenging, particularly in the presence of macular oedema, which can distort OCTA images. This study introduces an image-processing pipeline to evaluate DMI in patients with DMO using OCTA. Using a 3-dimensional methodology, the pipeline quantitatively assesses macular vascular perfusion, accounting for segmentation errors caused by macular oedema. Imaging data from 35 people with DMO and variable degrees of DMI were imaged using three OCTA devices (Heidelberg Spectralis, Optovue Angiovue, and Topcon Triton) and analysed using this 3D methodology. Key metrics including vessel density, skeletonized vessel density, average vessel radius, and our novel metric named Macular Vascular Volume (MVV), were extracted to assess consistency across two-time points, namely pre- and post-treatment with anti-Vascular Endothelial Growth Factor (anti-VEGF) agents and hence before and after resolution of macular oedema. Measurements with the novel process showed stability of the macular vessel metrics pre- and post-oedema resolution, compared with previously described OCTA metrics using current 2D segmentation methods. The proposed pipeline provides robust quantitative tools for evaluating DMI with DMO, with promise in clinical applications for improving diagnostic accuracy and optimising treatment outcomes.