The Impact of Image Processing Algorithms on Optical Coherence Tomography Angiography Metrics and Study Conclusions in Diabetic Retinopathy

• DOI: 10.1167/tvst.11.9.7
• Publication Date: 2022-09-15T14:33:33Z
• Authors: Isaac G. Freedman; Emily Li; Lucy Hui; Ron A. Adelman; Kristen Nwanyanwu; Jay C. Wang
• Keywords: Adaptive histogram equalization; Image subtraction

The purpose of this study was to evaluate the impact image processing on quantitative metrics in optical coherence tomography angiography (OCTA) images and conclusions patients with diabetes.This a single center, retrospective cross-sectional study. OCTA imaging Cirrus HD-OCT 5000 AngioPlex diabetes performed. 8 × mm superficial slab underwent 4 different preprocessing methods (none, background subtraction [BGS], foveal avascular zone brightness adjustment, contrast limited adaptive histogram equalization [CLAHE]) followed by binarization algorithms (global Huang, global Otsu, local Niblack, Phansalkar) ImageJ. Vessel density (VD), skeletonized VD (SVD), fractal dimension (FD) were calculated. Mixed-effect multivariate linear regressions performed.Two hundred eleven scans from 104 included. Of these scans, 67 (31.8%) had no diabetic retinopathy (DR), 99 (46.9%) nonproliferative DR (NPDR), 45 (21.3%) proliferative (PDR). Forty-eight 211 (22.7%) macular edema (DME). method used significantly impacted values VD, SVD, FD (all P -values < 0.001). On analysis, changed clinical variables associated FD. However, BGS CLAHE yielded more consistent significant covariates across multiple algorithms.The can any given analyzing metrics. Thus, caution is urged interpretation such studies. Background or may play role standardization processing.This work proposes strategies achieve robust analysis imaging, which especially important for trials.