Abstract
Purpose :
Recent advances in OCT enable high-resolution imaging of the retina to understand a variety of retinal disorders. However, systematic analysis of multiple OCT datasets of large cohorts is complex, time-consuming, and yet challenging. This study demonstrates a new methodology of visual analysis (VA) for comparing and automatically measuring volumetric OCT data of intra-retinal layers in children with Type 1 Diabetes mellitus (T1DM) exemplifying the features of VA.
Methods :
Unilateral eyes of 17 children with T1DM without diabetic retinopathy and of 31 healthy children were examined using SPECTRALIS OCT. High-resolution macular volume scans were obtained and analyzed using VA. Automated algorithms were applied to enable intra-layer segmentation and generation of deviation maps (DM) based on thickness maps (TM) of the experimental and control groups. Interactive visualization was used to explore the data and to selectively measure noticeable differences between the experimental group and the control group (C1), subgroups (duration of disease, glycemic level, mode of treatment) and the control group (C2), and individual subjects to the control group (C3). ETDRS-grid based central (C), pericentral (PC) and peripheral (P) regions of DMs were analyzed.
Results :
DMs of C1 demonstrated diffuse thinning, relatively greater at C and superior PC regions measuring 10.17±3.05µm and 7.33±2.9µm respectively for the total retina. Retinal nerve fiber layer (RNFL), ganglion cell layer (GCL) and inner plexiform layer showed mild to moderate thinning at different areas, although the extent and amount were greater in RNFL (PC: 1.98±1.46µm, P: 3.51±4.76 µm). DMs of C2 demonstrated localized thinning of GCL, greater in subgroups of longer duration of disease (C: 2.48±1.77µm, PC: 1.83±1.93µm) and subcutaneous insulin infusion mode of treatment (C: 5.29±3.52µm, PC: 2.11±4.08µm). The DMs of C3 demonstrated different patterns of thickness changes varying between subjects and between layers.
Conclusions :
VA generates TMs for all segmented retinal layers from an OCT dataset by measuring every single data point on the map. The derived DMs are of equally high spatial precision and are highly useful to readily investigate the subtle and localized thickness changes between groups. Overall, VA eases in-depth analysis of OCT data from cohorts and is effective in detecting thickness changes in complex multidisciplinary studies in an intuitive way.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.