Objective: To assess the reliability and validity of gray level co-occurrence matrices (GLCM) in the quantification of choriocapillaris and describe GLCM features in normal and eyes with resolved acute posterior multifocal placoid pigment epitheliopathy (APMPPE) and serpiginous choroiditis (SC).
Methods: In this, multicenter, reliability, validity and comparative study; OCTA was performed on eyes with resolved APMPPE and SC and normal individuals. CC texture classification, low flow area measurements and GLCM feature extraction were performed.
Results: A total of 13 normal, 8 APMPPE and 15 SC eyes were analyzed. All GLCM parameters demonstrated an excellent reliability. GLCM parameters were differently distributed across the three groups. Decision-tree based on the random forest predictive model provided an overall accuracy of 86% in classifying the three groups using GLCM features.
Conclusion: These data demonstrated an excellent reliability and validity of GLCM features in quantifying the choriocapillaris in healthy and diseased eyes.
Keywords: Acute posterior multifocal placoid pigment epitheliopathy; Uveitis; choriocapillaris; choroiditis; optical coherence tomography angiography; primary inflammatory choriocapillaropathies; quantitative analysis of choriocapillaris; serpiginous choroiditis.