Table 2_Retinal vessel metric analysis of type 1 diabetes mellitus in OCT angiography.docx

PurposeTo investigate retinal vascular characteristics in type 1 diabetes mellitus (T1DM) patients at different stages via optical coherence tomography angiography (OCTA).
DesignA retrospective observational study.
Method63 T1DM patients (110 eyes) who underwent OCTA (3*3 mm) examination and 40 age-matched healthy individuals (79 eyes) were included. A deep learning model was used to segment the retinal arteries and veins on OCTA images, and the vascular metrics in the macular area, including fractal dimension (FD), vessel diameter index (VDI), vascular length fraction (VLF), vascular tortuosity, and vessel density (VD) in different zones (fovea, superior parafoveal, inferior parafoveal, nasal parafoveal, and temporal parafoveal) were calculated.
ResultsIn accordance to the diagnostic criteria for diabetic retinopathy (DR), T1DM patients were classified into groups. There were 12 individuals (19 eyes) in the NPDR group, all of whom exhibited non-proliferative DR, and 51 (91 eyes) in the NDR (non-DR) group. The NDR group was divided into 2 subgroups according to the duration of T1DM, with 28 people (49 eyes) having a duration of less than 5 years and 23 people (42 eyes) having a duration of 5 years or more. Built-in metrics of OCTA showed that the VD in each zone of NPDR group was significantly lower than that of control group and NDR group (all p < 0.05). In addition, the VD of the inferior parafoveal vein (P = 0.022) and superior parafoveal artery (P = 0.03) were significantly decreased in the NPDR group. Comparing the VD among NDR subgroups and control group, the VD of the superficial inferior parafovea of T1DM patients of early stage (less than 5 years) was significantly lower than that of normal people (p < 0.05), while the VD of the inferior parafoveal artery increased (P = 0.012).
ConclusionWe employed a deep learning vessel segmentation model to analyze the changes in arterial and venous metrics in OCTA images of T1DM patients. Early damage of NPDR to large vessels occurs in the inferior parafoveal vein and the superior parafoveal artery. In patients without DR, the arterial VD of the inferior parafovea has a compensatory increase.

研究目的：借助光学相干断层扫描血管成像（optical coherence tomography angiography, OCTA），探究不同病程阶段1型糖尿病（type 1 diabetes mellitus, T1DM）患者的视网膜血管特征。
研究设计：回顾性观察性研究。
研究方法：纳入接受3×3mm范围光学相干断层扫描血管成像检查的63例1型糖尿病患者（共110只眼），以及40名年龄匹配的健康志愿者（共79只眼）。采用深度学习模型对光学相干断层扫描血管成像图像中的视网膜动静脉进行分割，并计算黄斑区不同分区（黄斑中心凹、旁黄斑上区、旁黄斑下区、旁黄斑鼻侧区、旁黄斑颞侧区）的血管参数，包括分形维数（fractal dimension, FD）、血管直径指数（vessel diameter index, VDI）、血管长度占比（vascular length fraction, VLF）、血管迂曲度以及血管密度（vessel density, VD）。
研究结果：依据糖尿病视网膜病变（diabetic retinopathy, DR）诊断标准，将1型糖尿病患者分为两组：非增殖性糖尿病视网膜病变（non-proliferative DR, NPDR）组12例（19只眼），均确诊为非增殖性糖尿病视网膜病变；非糖尿病视网膜病变（non-DR, NDR）组51例（91只眼）。根据1型糖尿病病程将NDR组进一步分为2个亚组：病程不足5年亚组28例（49只眼），病程≥5年亚组23例（42只眼）。光学相干断层扫描血管成像内置参数分析显示，NPDR组各分区的血管密度均显著低于对照组与NDR组（均p<0.05）；此外，NPDR组旁黄斑下区静脉血管密度（P=0.022）与旁黄斑上区动脉血管密度（P=0.03）均显著降低。对比NDR亚组与对照组的血管密度发现，早期（病程不足5年）1型糖尿病患者的浅层旁黄斑下区血管密度显著低于健康人群（p<0.05），但其旁黄斑下区动脉血管密度则出现代偿性升高（P=0.012）。
研究结论：本研究采用深度学习血管分割模型，分析1型糖尿病患者光学相干断层扫描血管成像图像中的动静脉参数变化。结果显示，非增殖性糖尿病视网膜病变早期对大血管的损害主要累及旁黄斑下区静脉与旁黄斑上区动脉；在无糖尿病视网膜病变的1型糖尿病患者中，旁黄斑下区动脉血管密度可出现代偿性升高。