Purpose
To investigate the use of ratiometric analysis of retinal layer thicknesses as an early indicator of diabetes.
Methods
Eight patients were chosen for this study. Four of the patients were clinically diagnosed to have diabetes (DM+) and four were used as control subjects (DM-). All patients’ retinas were imaged using optical coherence tomography (RTVue, Optovue, Inc.) and showed no morphological abnormalities. Retinal images were then manually segmented at selected layers (see fig. 1) and ratios between various layer thicknesses were compared between DM- and DM+ cases. After analysis, these ratios were correlated to the presence or absence of diabetes.
Results
It was found that a ratiometric analysis of retinal layer thicknesses in the foveal region highlighted many ratios in which the mean of the DM+ group differed from the mean of the DM- group by at least one standard deviation. Figure 2 shows 12 ratios taken from the foveal region in which both means lie outside of their respective standard deviations, indicating a statistically significant difference. While there is some interdependence between these ratios, they include contributions from every segmented layer. A physiologically significant correlation is the fact that all of these ratios involve the inner segment and almost half of them involve the retinal pigment epithelium, two layers that would be directly affected by vascular changes associated with diabetic retinopathy.
Conclusions
A ratiometric analysis on layer thicknesses in the retina of diabetic and control patients reveals a statistically different set of thickness ratio values for diabetic patients. These results are encouraging for future efforts to diagnose or screen for diabetes at an early stage, before symptoms or gross retinal abnormalities occur. Previous studies have investigated layer thicknesses in the retina to draw conclusions about the presence of diabetes; however, an analysis of the ratios of these layer thicknesses is a more robust measure. Variations in overall retinal thickness between patients will not affect the accuracy of ratio measurements.
Keywords: 499 diabetic retinopathy •
552 imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) •
550 imaging/image analysis: clinical