Fig 1.
Differences among diabetic eyes that are not fully characterized by retinal thickness, demonstrated with cross-sectional views and average central macular thickness from an iVue SD-OCT.
(A) Typical diabetic macular edema, showing a thickened retina that results in a high value of central macular thickness. There are fluid-filled cystic spaces that appear dark, hyper-reflective lipid and protein deposits (hard exudates), and disrupted photoreceptor layers that lie beneath retina blood vessels. The large areas of fluid produce multiply scattered light, instead of leading to interference, and therefore appear dark. (B) Significant pathological changes, but a low value of retinal thickness typically attributed due to damage to neurons and their support cells. There are numerous hard exudates, and significant disruption to retinal layers including photoreceptors. (C) Traction along with a detaching vitreous, the topmost reflective layer that is tilted, leading to retinal thickening, but not very severe diabetic changes. The large black areas within the retina demonstrate the fluid built up by the traction and are consistent with a high value of central macular thickness. (D) Diabetic retina with normal retinal thickness and minimal diabetic changes. These data are from the large dataset collected in a group of clinics for the underserved in Alameda County, CA, as described in ref [4].
Fig 2.
The five domains where thickness and frequency content were measured.
A) The full retinal domain. B) The domain from the ISOS junction to the ILM. C) The domain from the ISOS to the boundary between the IPL and INL. D) The domain from the ISOS junction to the boundary between the NFL and GCL. E) The domain from the boundary between the RPE and CH to the ISOS junction.
Fig 3.
A) Representative b-scan from a diabetic subject, showing the full retinal domain, with the areas outside of the retina being set to the average intensity of the retina, and showing the ramping effects for edge smoothing on the sides of the retina. B) B-scan of the control subject. C) Average pixel intensity through the retina, with 0 as the inner-most position of the retina for the diabetic and control subjects in A and B, with the diabetic in blue and control in orange. D) Log transform of the power spectra for the subjects in A and B.
Fig 4.
Thickness measurements for the domain of the full retina.
Location 0 is the fovea, negative locations are in degree steps temporal to the fovea, and positive locations are in degree steps nasal to the fovea. There were no statistically significant differences in thickness between the diabetic and control subjects at any location, with diabetics in blue and controls in orange. That is, the trends seen in the figure are not significant, for diabetic subjects on average being thicker at all locations temporal to the fovea, and the control subjects being thicker on average at the fovea and for all locations nasal to the fovea.
Table 1.
Retinal thicknesses and standard deviations for ETDRS regions.
Fig 5.
Thickness measurements of the domain ranging from the boundary between the RPE and CH to the ISOS junction.
Location zero is the fovea, negative locations indicate degrees temporal to the fovea, and positive locations indicate degrees nasal to the fovea. Orange circles are the control subjects, and blue circles are the diabetic subjects. Control subjects were significantly thicker than the diabetic subjects at 1 deg temporal to the fovea and 3 deg nasal to the fovea. These data do not support the idea that diabetic subjects as a whole develop thicker deep retinal layers prior to clinical disease, for locations in the central retina.
Fig 6.
A) Average power spectrum for diabetics and controls for the domain that ranges from the boundary between the RPE and CH to the ISOS junction. B) Controls had statistically significant more power in the frequency range from 25.5 to 29 microns/cycle compared to diabetics. C) Diabetics had statistically significant more power in the frequency range from 15.5 to 18.2 microns/cycle compared to controls.
Fig 7.
A) Individual z-scores for the control subjects computed for the power at each frequency for the domain from the boundary between the RPE and CH to the ISOS junction. B) Individual z-scores for the diabetic subjects in the same domain. C) The averaged z-score for the diabetic subjects at each frequency for the same domain. The diabetics consistently had more power at the higher frequencies than the control subjects.
Table 2.
Linear regressions and R2 values for age vs ETDRS region.