Abstract
Purpose :
This case series demonstrates a proof of concept using automated machine learning to quantify the change in limbal episcleral vessel morphology in glaucoma patients who undergo trabeculotomy ab interno (Trabectome). Currently there are no reliable noninvasive methods to quantify changes in anatomy after minimally invasive glaucoma surgeries (MIGS). Intraocular pressure (IOP) and gonioscopy remain the two parameters that surgeons can assess for MIGS efficacy. One type of MIGS, Trabectome, is thought to decrease IOP by decreasing aqueous humor outflow resistance. Sclera images may offer a methodology for assessing changes seen noninvasively as a result of Trabectome.
Methods :
This study retrospectively analyzed 3 eyes from 2 patients who underwent combined Trabectome with phacoemulsification cataract surgery by 1 surgeon in a hospital setting. Images were taken immediately before and 1 month after the procedure using the Zeiss IOLMaster 500. The open-source image processing package Fiji (ImageJ mirror) was used to load images that were cropped to rectangles of the sclera-only nasal region. The Fiji plug-in Trainable Weka Segmentation was used with 4 reference patients’ images from a different dataset to train a model to identify scleral vessels from sclera. The trained model was then applied to study images to calculate the % area of scleral vessels. 3 rectangles of each patient’s eye were cropped from the nasal region to optimize either width, height, or most salient vessels, and averages and standard deviations were calculated using Microsoft Excel 2021.
Results :
The first patient’s right eye had an average % area of scleral vessels compared to white sclera of 9.10±0.27% pre-procedure, and 8.72±0.30% post-procedure. The second patient’s right eye had an average % area of 13.75±0.83% pre-procedure compared to 7.18±0.72% post, and left eye average of 12.77±1.21% pre-procedure compared to 11.05±0.65% post. All three eyes had greater than 30% IOP reduction after Trabectome.
Conclusions :
Overall there was a decrease in visible sclera vessel area across all three patients who underwent the Trabectome surgery. This presumably is due to less blood and more aqueous in these episcleral vessels following Trabectome. The images captured noninvasively with the IOLMaster 500 along with the machine learning model may be useful adjunctive tools to assess surgical outcomes following MIGS.
This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.