Abstract
Purpose :
Evaluate the predictive ability of a deep learning-based, automated, macular fluid segmentation algorithm to determine long-term visual acuity (VA) outcomes in wet age-related macular degeneration (wARMD) patients using baseline swept-source optical coherence tomography (SS-OCT) and OCT-angiography (OCT-A) data.
Methods :
Twenty-two SS-OCT volumes of the macula, comprising 5,632 images from 22 wARMD subjects were used to assess retinal layer thicknesses, quantify intraretinal fluid (IRF), subretinal fluid (SRF) and fluid in serous pigment epithelium detachments (PED). Layer thicknesses were manually corrected and fluid segmentation was performed using a novel, deep learning algorithm with results validated relative to two expert graders. OCT-A data was used to manually define the extent of the choroidal neovascularization (CNV) in each scan. Patients received 2 mg of intravitreal aflibercept injections monthly for 3 months, then bimonthly for 12 months. Baseline OCT morphological features and measurements were correlated using the Pearson correlation coefficient (PCC) to changes in VA to determine which features impacted the long-term visual outcomes.
Results :
Total fluid volume at baseline and change in LogMar at week 52 relative to baseline had the closest correlation (PCC=0.652, p=0.005). Fluid was subsequently sub-categorized into IRF, SRF and PED, with PED volume having the next highest correlation (PCC=0.648, p=0.005). Average total retinal thickness in isolation gave a lower correlation (PCC=0.334, p=0.189), and mean CNVM size (um2) from 3 mm OCT-A scans was very low (PCC=0.072, p=0.784). When two features were combined and correlated with visual outcomes, the best correlation increased to PCC=0.695 (p=0.002) using mean CNVM size and total fluid volume.
Conclusions :
In isolation, total fluid volume best correlates with change in LogMar values between baseline and week 52. In combination with complimentary information from OCT-A, an improvement in the linear correlation score was observed. Average total retinal thickness provided a lower correlation and thus provides a lower predictive outcome than other metrics assessed. Clinically, a machine-learning approach to analyzing fluid metrics combined with lesion size may provide an advantage in personalizing therapy predicting VA outcomes.
This is a 2021 ARVO Annual Meeting abstract.