OCTA and OCT Image Acquisition and Quality Assessment in the Framingham Heart Study

Abstract

Purpose : To describe the design and execution of an optical coherence tomography (OCT) and angiography (OCTA) protocol in the Framingham Heart Study (FHS).

Methods : A prospective, IRB-approved study was conducted from 2020-2022 among the Offspring Generation (Exam 10) and OMNI-1 (Exam 5) cohorts (Framingham, MA). Foveal-centered 3x3 mm² OCTA, 200x200 Optic Disc Cube OCT, and Axial Length (AL) were acquired in quadruplicates on both undilated eyes (starting on the right eye) of each subject using commercial instruments. Imaging technicians without previous ocular imaging experience were trained using both remote and in person training protocols. All imaging procedures were completed in ~15 minutes. Image Quality Control (QC) was quantitatively assessed as “Excellent”, “Good”, “Suboptimal”, or not acquired by consensus of two trained graders not involved in image acquisition and based on the following criteria: signal strength, motion artifacts, medial opacities, focus, and decentration. The single best image was identified for future analysis. McNemar’s test was used for statistical comparisons.

Results : 959 subjects were recruited with the following characteristics: Age 76 ±7, Women 561 (59%), taking anti-hypertensives 525 (56%), Diabetes 99 (11%), Hyperlipidemia 63 ±19, Prevalent cardiovascular disease 96 (10%). Complete OCT, OCTA and AL data were acquired on 866/959 (90%), 923/959 (96%), and 859/959 (89%) subjects, respectively. Lack of gradable images was due to small pupil size, poor fixation, time constraints or hardware problems. Image QC assessed OCTA images as: 734/1918 (38%) “Excellent”, 533/1918 (28%) “Good”, and 471/1918 (25%) “Suboptimal” with 180/1918 (9%) not acquired. Image QC assessed OCT images as: 1299/1918 (68%) “Excellent”, 346/1918 (18%) “Good”, 162/1918 (8%) “Suboptimal”, with 111/1918 (6%) not acquired. (Table 1) There was a significant difference between frequency of unsuccessful image acquisitions for left compared to right eyes for both OCTA(p<0.01) and OCT(p<0.01).

Conclusions : We report preliminary data demonstrating favorable results for image acquisition and rating in a large cohort of non-dilated eyes. We show a significant difference in image acquisition, likely due to the order the images were acquired. Further analysis of variables impacting data collection and quantifiable metrics will be helpful in refining and improving future image collection efforts.

This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.