Dabir Supriya, Bhatt Vaidehi, Bhatt Deepak, Rajan Mohan, Samant Preetam, Munusamy Sivakumar, Webers C A B, Berendschot T T J M
Department of Retina, Rajan Eye Care Pvt Ltd, Chennai, India.
Rajiv Gandhi Medical College, Thane, India.
PLoS One. 2020 Dec 31;15(12):e0244828. doi: 10.1371/journal.pone.0244828. eCollection 2020.
To compare the characteristics of eyes that had manual vs. automated segmentation of choroidal neovascular membrane (CNVM) using optical coherence tomography angiography (OCTA).
All patients with CNVM underwent OCTA using the Zeiss Angioplex Cirrus 5000. Slabs of the avascular outer retina, outer retina to choriocapillaris (ORCC) region and choriocapillaris were generated. Manual segmentation was done when there were significant segmentation artifacts. Presence of activity of CNVM was adjudged by the presence of subretinal fluid (SRF) on structural OCT and was compared to activity detected on en face OCTA slabs based on well-defined criteria.
Eighty-one eyes of 81 patients were recruited of which manual segmentation was required in 46 (57%). Eyes with automated segmentation had significantly more CNVM in the ORCC (75%) whereas those with manual segmentation had deeper CNVM (sub-RPE = 22%, intra-PED = 22%) (p<0.001). Twenty eyes (25%) were found to have active CNVM on both the structural OCT and OCTA while an additional 19 eyes were presumed to have active CNVM on OCTA alone. There was only modest concordance between disease activity detected using structural OCT and OCTA (Kappa = 0.47, 95% CI = 0.30 to 0.64).
Manual segmentation of OCTA is required in more than 50% eyes with CNVM and this progressively increases with increasing depth of CNVM location from the ORCC to below the RPE. There is moderate concordance between OCTA and structural OCT in determining CNVM activity.
比较使用光学相干断层扫描血管造影（OCTA）对脉络膜新生血管膜（CNVM）进行手动分割与自动分割的眼部特征。
所有患有CNVM的患者均使用蔡司Angioplex Cirrus 5000进行OCTA检查。生成无血管外层视网膜、外层视网膜至脉络膜毛细血管（ORCC）区域和脉络膜毛细血管的层面图像。当存在明显的分割伪影时进行手动分割。根据结构OCT上视网膜下液（SRF）的存在来判断CNVM的活性，并根据明确的标准将其与在OCTA层面图像上检测到的活性进行比较。
招募了81例患者的81只眼，其中46只眼（57%）需要手动分割。自动分割的眼中，ORCC区域的CNVM明显更多（75%），而手动分割的眼中，CNVM更深（视网膜色素上皮下 = 22%，息肉状脉络膜血管病变内 = 22%）（p<0.001）。20只眼（25%）在结构OCT和OCTA上均被发现有活动性CNVM，另外19只眼仅在OCTA上被推测有活动性CNVM。使用结构OCT和OCTA检测到的疾病活性之间只有适度的一致性（Kappa = 0.47，95% CI = 0.30至0.64）。
超过50%的CNVM患眼需要对OCTA进行手动分割，并且随着CNVM位置从ORCC到视网膜色素上皮下方深度的增加，这种情况会逐渐增多。在确定CNVM活性方面，OCTA和结构OCT之间存在中度一致性。
