Quantitative assessment of textural features in the early detection of diabetic retinopathy with optical coherence tomography angiography

The diagnostic power of the textural features and fractal dimension (FD) for early detection of diabetic retinopathy (DR) were quantitatively evaluated with optical coherence tomography angiography (OCTA). 41 normal healthy OD eyes (age: 46.41 ± 13.69), and 10 diabetic OD eyes (age: 60.90 ± 13.46) in the early stages of DR (mild or moderate non-proliferative diabetic retinopathy (NPDR)) were employed. Four retinal vascular plexuses including nerve fiber layer vascular plexus (NFLVP), superficial vascular plexus (SVP), intermediate capillary plexus (ICP) and deep capillary plexus (DCP) were generated by using OCTA. The FD and the five textural features including contrast, correlation, entropy, energy, and homogeneity in four parafoveal sectors were extracted from OCTA images. The factor of aging on textural features and FD was evaluated based on the comparisons among five normal healthy subgroups. Our results showed that FD in superior sector of NFLVP and in nasal sector of ICP had the significant decreases when comparing the older healthy subgroup (age range: 60–69) with the younger healthy subgroup (age range: 20–29). Our results also indicated that the correlation did not show the significant differences in all sectors of the four retinal sublayers among the normal healthy subgroups except in the temporal sector in NFLVP. Furthermore, our results indicated that the correlation in nasal and inferior sectors in SVP can effectively differentiate diabetic patients in early stages of DR from normal healthy subjects with the highest AUROC values. In our study, the specific textural feature - correlation can effectively stage the early DR, which may contribute to the diagnosis of DR in clinic practice.