Automated Quantitation of Choroidal Neovascularization: A Comparison Study Between Spectral-Domain and Swept-Source OCT Angiograms

• Published in: Investigative ophthalmology & visual science Vol. 58; no. 3; pp. 1506 - 1513
• Main Authors: Zhang, Qinqin, Chen, Chieh-Li, Chu, Zhongdi, Zheng, Fang, Miller, Andrew, Roisman, Luiz, Rafael de Oliveira Dias, Joao, Yehoshua, Zohar, Schaal, Karen B., Feuer, William, Gregori, Giovanni, Kubach, Sophie, An, Lin, Stetson, Paul F., Durbin, Mary K., Rosenfeld, Philip J., Wang, Ruikang K.
• Format: Journal Article
• Language: English
• Published: United States The Association for Research in Vision and Ophthalmology 01.03.2017
• Subjects: Aged; Aged, 80 and over; Algorithms; Automation; Choroid - blood supply; Choroid - pathology; Choroidal Neovascularization - pathology; Female; Fluorescein Angiography - methods; Fundus Oculi; Humans; Male; Middle Aged; Multidisciplinary Ophthalmic Imaging; Prospective Studies; Reproducibility of Results; Tomography, Optical Coherence - methods
• ISSN: 1552-5783; 0146-0404; 1552-5783
• DOI: 10.1167/iovs.16-20977

To compare the lesion sizes of choroidal neovascularization (CNV) imaged with spectral-domain (SD) and swept-source (SS) optical coherence tomography angiography (OCTA) and measured using an automated detection algorithm. Patients diagnosed with CNV were imaged by SD-OCTA and SS-OCTA systems using 3 × 3-mm and 6 × 6-mm scans. The complex optical microangiography (OMAGC) algorithm was used to generate the OCTA images. Optical coherence tomography A datasets for imaging CNV were derived by segmenting from the outer retina to 8 μm below Bruch's membrane. An artifact removal algorithm was used to generate angiograms free of retinal vessel projection artifacts. An automated detection algorithm was developed to quantify the size of the CNV. Automated measurements were compared with manual measurements. Measurements from SD-OCTA and SS-OCTA instruments were compared as well. Twenty-seven eyes from 23 subjects diagnosed with CNV were analyzed. No significant differences were detected between manual and automatic measurements: SD-OCTA 3 × 3-mm (P = 0.61, paired t-test) and 6 × 6-mm (P = 0.09, paired t-test) scans and the SS-OCTA 3 × 3-mm (P = 0.41, paired t-test) and 6 × 6-mm (P = 0.16, paired t-test) scans. Bland-Altman analyses were performed to confirm the agreement between automatic and manual measurements. Mean lesion sizes were significantly larger for the SS-OCTA images compared with the SD-OCTA images: 3 × 3-mm scans (P = 0.011, paired sample t-test) and the 6 × 6-mm scans (P = 0.021, paired t-test). The automated algorithm measurements of CNV were in agreement with the hand-drawn measurements. On average, automated SS-OCTA measurements were larger than SD-OCTA measurements and consistent with the results from using hand-drawn measurements.