Abstract
Purpose :
Thickness profiles of intraretinal layers serve as a biomarker to detect neurodegenerative changes in the retina. Our recent advancement in Visual Analytics (VA) software [1] significantly reduces the complexity of analyzing large OCT volume datasets while promoting a precise detection of thickness changes in retinal layers. Here, we aimed at evaluating deviation maps obtained from VA in relation to grid-based results from conventional analysis (CA) in patients with type 2 diabetes mellitus (T2DM).
Methods :
OCT was performed on unilateral eyes of 33 patients with T2DM without diabetic neuropathy and 22 controls to obtain macular volume scans. Segmented retinal nerve fiber layer (RNFL), ganglion cell layer (GCL), inner plexiform layer (IPL), inner nuclear layer (INL) and total retina (TR) were analyzed using VA and CA. CA: ETDRS grid based averaged thickness values from 9 sectors representing the foveal center, inner macula and outer macular regions were compared between groups. VA: Every single point of OCT data was measured to generate thickness maps (TMs) for all evaluated layers. Color-coded deviation maps (DevMs) were generated by compiling and comparing TMs of the patient and control groups. Abnormal areas of significant changes were identified directly on DevMs.
Results :
CA: In patients, the TR, GCL and IPL demonstrated significant thinning (mean±SD; µm) in all quadrants of inner (-10.5±1.0; -5.8±0.3; -2.8±0.3; p<0.05) and outer (-10.8±0.4; -3.5±0.7; -2.0±0.73; p<0.05) macular regions. Thickening of foveal center (+3.6±0; p<0.05) was noted only in INL. VA: In contrast to CA, thickness changes were directly explored on color-coded DevMs. DevMs of TR, GCL, and IPL exhibited a pattern of unaffected foveal center surrounded by large areas of thinning at the inner, and outer macular regions. For RNFL, besides temporal thickening demonstrated by CA, an additional thinning in the nasal quadrant was noted. Overall, the RNFL and INL showed minimal changes with small and localized areas of thinning and thickening.
Conclusions :
VA based DevMs provide a topographical overview of retinal layer changes. In comparison to grid-based CA, DevMs are spatial specific and highly precise in detecting preclinical retinal thickness changes. Hence, combining CA and VA will improve the analysis of retinal thickness data.
[1] Roehlig et al. Visual Computer 34(9), 1209-1224.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.