Validation of Automated Software for Guttae Quantification in Fuchs Endothelial Corneal Dystrophy

Purpose: Corneal guttae are the hallmark pathologic phenotype of Fuchs endothelial corneal dystrophy (FECD). Tools performing quantitative analysis of guttae and their changes over time are needed as medical therapies advance. We developed and validated a novel automated software platform that reports the guttae area ratio (GAR) in FECD specular microscopy images. Methods: Patients with FECD with modified Krachmer grade (mKg) 2 to 5 disease and no prior keratoplasty were enrolled and imaged prospectively with one specular microscope unit. In a validation study, expert manual grading and automated software grading of GAR were compared for agreement. Additional images were analyzed to correlate automated GAR with disease grade and subclinical edema by Scheimpflug imaging. Results: Nineteen images were included in the validation study. Mean manual GAR (0.400 ± 0.213) and mean automated GAR (0.395 ± 0.182) demonstrated equivalence (P = 0.003). Additionally, 446 images (109 eyes) demonstrated that automated GAR increases with worsening disease in mKg 2 to 5. Most mKg 4 images (n = 172, 93.5%; 49 eyes) and mKg 5 images (n = 59, 55.1%; 19 eyes) had no evidence of definite subclinical edema. Repeat image analysis demonstrated high repeatability of GAR (intraclass correlation = 0.85; 95% confidence interval, 0.78-0.90). Conclusions: This study validates that automated software for guttae quantification in FECD specular microscopy images is equivalent to manual expert grading. GAR, an objective metric that increases in direct proportion to mKg, may precede development of subclinical edema. Translational Relevance: This software supports precise quantification of FECD disease secondary to guttae, allowing for objective measurement of disease progression and therapeutic efficacy.