Abstract
Purpose :
Current optical coherence tomography (OCT) image-analysis tools for AMD diagnosis that rely on thickness or volume measurements have limited accuracy as they are vulnerable to physiological variation. We optimized sinuosity, a ratio measure used in environmental science to quantify river curvature, to assess outer retinal curvature and tested this novel biomarker for intermediate AMD (iAMD) identification.
Methods :
Forty eyes from 20 consecutive iAMD and 20 non-AMD patients were recruited. Inclusion criteria were age over 50 years, a macular OCT volume scan (Spectralis OCT) with a quality score > 15 and confirmed clinical diagnosis of iAMD or non-AMD. Retinal pigment epithelium (RPE) and Bruch’s membrane (BM) length within a 1000µm vertical extent around the fovea was measured using custom MATLAB code and sinuosity calculated as RPE:BM length. Sinuosity was determined for dense (120µm between scans; n=9 scans) versus sparse (240µm between scans; n=5 scans) B-scan density, and whole (4.4-10.3mm) versus cropped (central 1000um) B-scans. The primary outcome measure was the area-under-the-receiver-operator-curve (aROC).
Results :
Sinuosity was significantly greater for iAMD (range: 1.034-1.205) than non-AMD patients (range: 0.986-1.017) under all conditions tested (p=0.001-0.016). In iAMD, scan length (p=0.011-0.051) significantly affected sinuosity values whereas B-scan density (p=0.208-0.845) did not. Neither factor significantly impacted sinuosity in non-AMD patients. Mean, median and maximum sinuosity may capture different characteristics of AMD used to monitor progression such as drusen load and maximum drusen size. A comparison of index test accuracy from using these values yielded aROC results of 0.890-0.958 with best performance obtained using the mean sinuosity of cropped B-scans (0.958). The mean sinuosity of cropped scans remained significantly different between groups after age and B-scan length adjustment (p<0.001).
Conclusions :
We found the diagnostic accuracy of using outer retinal curvature quantified as RPE sinuosity for iAMD identification promising. The optimal protocol was mean RPE sinuosity from the central 1000um of multiple scans with an interscan spacing of 240µm.
This abstract was presented at the 2023 ARVO Imaging in the Eye Conference, held in New Orleans, LA, April 21-22, 2023.