Murakawa Sakura, Maruko Ichiro, Kawano Taizo, Hasegawa Taiji, Iida Tomohiro
Department of Ophthalmology, Tokyo Women's Medical University, 8-1 Kawadacho, Shinjuku-ku, Tokyo, 162-8666, Japan.
Graefes Arch Clin Exp Ophthalmol. 2019 Jun;257(6):1119-1125. doi: 10.1007/s00417-019-04275-5. Epub 2019 Feb 19.
To determine the effects of averaging five en face optical coherence tomography angiographic (OCTA) images on the quality of the images in eyes with a choroidal neovascularization (CNV).
Twenty-seven eyes of 25 patients (18 men, 7 women; average age 71.0 years) with a CNV were examined by OCTA (OCT HS-100, Canon. Japan). A 3 × 3-mm image including the CNV was recorded and automatically segmented between the retinal outer layers. Analyses were performed on a single image (S-image) and the average of five single images of the same area (A-images). The region of the CNV was selected by ImageJ, and the peak signal-to-noise ratio (PSNR), the vascular density (VD), fractal dimension (FD), and the noise component using band pass filter (BPF) processing of the S- and A-images of each case were compared.
The average PSNR for the A-images was 14.0 which was significantly higher than the 12.2 for the S-images (P < 0.01). However, the average VD was 33.6% for the S-images and 34.8% for the A-images (P > 0.1). The average FD was 1.67 for the S-images and 1.54 for the A-images (P < 0.01). The mean luminance difference obtained by subtracting the luminance of the A-image from the S-image after BPF processing was 10.41 ± 14.66 db which was positive for all eyes.
The better quality of the A-images of a CNV and absence of a significant difference in the vascular density indicates that the improvement was due to the removal of the same signal levels of the noise component and blood vessels.
确定对脉络膜新生血管(CNV)眼的五幅表面光学相干断层扫描血管造影(OCTA)图像进行平均处理对图像质量的影响。
使用OCTA(日本佳能公司的OCT HS - 100)对25例患者(18例男性,7例女性;平均年龄71.0岁)的27只患有CNV的眼睛进行检查。记录一幅包含CNV的3×3毫米图像,并在视网膜外层之间自动分割。对单幅图像(S图像)和同一区域的五幅单幅图像的平均值(A图像)进行分析。通过ImageJ选择CNV区域,比较每种情况的S图像和A图像经带通滤波器(BPF)处理后的峰值信噪比(PSNR)、血管密度(VD)、分形维数(FD)以及噪声成分。
A图像的平均PSNR为14.0,显著高于S图像的12.2(P < 0.01)。然而,S图像的平均VD为33.6%,A图像的平均VD为34.8%(P > 0.1)。S图像的平均FD为1.67,A图像的平均FD为1.54(P < 0.01)。BPF处理后,从S图像中减去A图像的亮度得到的平均亮度差为10.41±14.66分贝,所有眼睛均为正值。
CNV的A图像质量更好且血管密度无显著差异,表明这种改善是由于去除了噪声成分和血管相同信号水平的结果。