Abstract
Purpose :
To develop an artificial inteligence based method to assess, quantify and categorized inflammation in the anterior chamber of the eye in DMEK patients using SS-OCT image analysis. To explore the correlation between SS-OCT based inflammation analyses with the clinical outcomes of the technique.
Methods :
Prospective, randomized clinical trial with 128 that underwent unilateral DMEK or triple procedure. An image processing algorithm was developed to process anterior OCT images acquired by Tomey CASIA-II. Mnimun particle size, brightness treshold and noise sensitivity were characterized. Number and relative size of particles, as well as background brightness of anterior chamber aqueous was extracted. Full anterior segment, non averaged, radial OCT images were acquired at day 0, 1, 3, 7, 30 and 90. Images were analysed and compared to the clinical findings and the SUN grading of flare and cells in the anterior chamber.
Results :
Number of particles per mm2 ranged from 0 to 15.95 (mean 9.8), with a mean estimated particle diameter 21um (SD=6.4). Distribution of particle size showed double peak configuration (peaks around 15-18um and 30-35um). Particle density was on average 1.0, 0.5, 0.8 and 1.5 per mm2 for clinical SUN scores 0, 0.5, 1, 2 and 3, respectively (Spearman correlation coefficient rho of 0.62). Particle density decrease with consequtive follow ups specially during the first postoperative week.
Conclusions :
Software based SS-OCT image analyses might be a feasible non-invasive tool for objective assessment of inflammation after keratopalsty with clear potential applications in the management of postoperative medication and the presvention of graft rejections.
This is a 2020 ARVO Annual Meeting abstract.