The preliminary study of diabetic retinopathy detection based on intensity parameters with optical coherence tomography angiography

• Published in: Kompʹûternaâ optika Vol. 47; no. 4; pp. 620 - 626
• Main Authors: Hou, J., Shi, H., Gao, W., Lin, P., Li, B., Shi, Y., Matveeva, I.A., Zakharov, V.P., Bratchenko, I.A.
• Format: Journal Article
• Language: English
• Published: Samara National Research University 01.08.2023
• Subjects: diabetic retinopathy; intensity; optical coherence tomography angiography; variance
• ISSN: 0134-2452; 2412-6179
• DOI: 10.18287/2412-6179-CO-1261

In this study, the diagnostic abilities of intensity parameters of optical coherence tomography angiography (OCTA) images in the early detection of diabetic retinopathy (DR) were determined. 78 normal healthy eyes, 10 diabetic eyes with mild non-proliferative diabetic retinopathy (NPDR), and 10 diabetic eyes with moderate NPDR were employed. Four retinal vascular plexuses were generated by using OCTA, which included the nerve fiber layer vascular plexus (NFLVP), superficial vascular plexus (SVP), intermediate capillary plexus (ICP) and deep capillary plexus (DCP). The parafoveal zone in each OCTA image was divided into four sectors which were the superior, temporal, inferior, and nasal sectors. Five intensity parameters including the mean, median, variance, skewness, and kurtosis of intensities were calculated for each sector. The factor of aging was evaluated among normal healthy subgroups. The diagnostic abilities of intensity parameters were evaluated between normal healthy subjects and diabetic patients with DR. Our results showed that the variance of intensities in superior sector in ICP achieved the highest AUROC value of 0.95 with the sensitivity of 0.87 and the specificity of 1.000 when comparing the diabetic patients with the mild NPDR to normal healthy subjects. The mean intensity in superior sector in ICP achieved the second highest AUROC value of 0.95 with the sensitivity of 0.90 and the specificity of 0.90 when comparing the diabetic patients with the moderate NPDR to normal healthy subjects. The proposed approach could offer a simple way to differentiate diabetic patients with early DR from normal healthy subjects without performing the relatively complicated image processing techniques.