We continue to cover some interesting posterior segment-related e-posters presented during The Royal Australian and New Zealand College of Ophthalmologists (RANZCO) 52nd Annual Scientific Meeting…
360-Degree Laser Retinopexy in Primary Vitrectomy
Single surgery anatomical success (SSAS) is a main outcome in primary rhegmatogenous retinal detachment (RRD) repair. As the risk of recurrent detachment and poor functional outcomes increase with subsequent procedures, the use of 360-degree prophylactic endolaser photocoagulation has been proposed to reduce re-detachment rates. Dr. Matthew Peters from The Royal Brisbane and Women’s Hospital (RBWH) investigated the use of 360-degree laser retinopexy in the treatment of RRD.
The study involved 190 RBWH patients who underwent primary pars plana vitrectomy (PPV) with or without scleral buckling (SB) from June 2017 to December 2020. Patients were divided into the 360-degree laser group (n=130) or limited laser retinopexy group (n=62).
The researchers found significant association of the 360-degree laser use with worse preoperative LogMAR visual acuity, male sex and higher grade of proliferative vitreoretinopathy (PVR). “360-degree laser was more likely to be applied in patients with more severe disease and worse prognosis and is part of more extensive operations. Even accounting for these factors, there is no significant difference in SSAS, final LogMAR visual acuity or complication rates between 360-degree laser and limited laser groups,” Dr. Peters said.
Correlation Between Retinal Ganglion Cell Layer Thickness
Retinal microvascular and structural changes may precede cognitive symptom onset in Alzheimer’s disease (AD). Using them as biomarkers may allow early identification of asymptomatic individuals at risk of AD, noted Dr. Samran Sheriff from Macquarie University, Sydney.
Dr. Sheriff and colleagues did a study to evaluate the correlations between retinal ganglion cell layer thickness (RGCL) and various domains of neuropsychological testing (Mini Mental State Exam [MMSE] and Symbol Digit Modality) in 95 healthy aging subjects recruited from the Optic Nerve Decline and Cognitive Change study.
“We found that the MMSE correlated with the Global RNFL (β = -4.409, p = 0.04), temporal superior RNFL (β = -11.2, p = 0.002), and naso-superior RNFL (β = -9.44, p = 0.002). The nasal superior region analysis also correlated with Symbol Digit Modality (β = 8.300, p = 0.002). This preliminary data supports the idea that the changes in retinal ganglion cell layer thickness correlate with specific neuropsychological domains,” he said, adding that future research will unravel whether a stronger statistical relationship between RGCL thickness and specific neuropsychological domains exists in dementia and psychiatric disorders.
Automated Detection of Glaucoma from Retinal Fundus Images using Different Fundus Cameras
Retinal fundus images are routinely used for diabetic photo screening and may also be used to screen for glaucoma. Nearly 30 retinal fundus cameras from different manufacturers are currently in use, such as the ZEISS VISUCAM 500, Canon CR-2 and Topcon TRC retina camera, according to Dr. James McKelvie from the University of Auckland, New Zealand. He and colleagues set out to evaluate the accuracy of automated glaucoma detection from retinal fundus images obtained with different fundus cameras.
The experiment was conducted in three steps: The researchers first cropped the images around the optic nerve head. Then, they extracted features using pre-trained deep neural network. Finally, they trained an AI model using a fundus camera and tested it with another fundus camera.
They noticed that an average accuracy of 96% was obtained when train and test images are from the same camera. However, the average accuracy dropped to 52% when train and test images are from two different fundus cameras. The authors concluded that while AI-based automated detection of glaucoma from fundal images is possible, AI models trained on one camera may have greatly reduced accuracy in grading images obtained from other cameras. Hence, they suggested that for clinical use, AI model accuracy needs to be reassessed for each camera type, and designing domain generalized glaucoma detection models will improve the robustness of these systems.
Editor’s Note: The 52nd Annual Scientific Congress of The Royal Australian and New Zealand College of Ophthalmologists (RANZCO Brisbane 2022) was held virtually from February 26 to March 1. Reporting for this story took place following the event.
