August 2014
Volume 55, Issue 8
Letters to the Editor  |   August 2014
Author Response: Retinal Segmentation to Demonstrate Hyperplasia in Ataxia of Charlevoix-Saguenay: Critique on Study Methodology and Results
Author Notes
Investigative Ophthalmology & Visual Science August 2014, Vol.55, 4729-4730. doi:
  • Views
  • PDF
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      E. Garcia-Martin; Author Response: Retinal Segmentation to Demonstrate Hyperplasia in Ataxia of Charlevoix-Saguenay: Critique on Study Methodology and Results. Invest. Ophthalmol. Vis. Sci. 2014;55(8):4729-4730. doi:

      Download citation file:

      © ARVO (1962-2015); The Authors (2016-present)

  • Supplements
We carefully read the letter by Albrecht and colleagues 1 and revised our article “Retinal Segmentation as Noninvasive Technique to Demonstrate Hyperplasia in Ataxia of Charlevoix-Saguenay.” 2 We have responded to each of the items outlined in their letter: 
With great interest we read the recent article by Garcia-Martin and colleagues 1 on the retinal segmentation of optical coherence tomography (OCT) scans from patients with autosomal recessive spastic ataxia of Charlevoix-Saguenay. We were impressed by the thorough ophthalmological and radiological workup of these patients with genetic confirmation of this rare disease. Therefore, it comes as a surprise that the quantitative results of the automated segmentation and the qualitative description of some of the retinal layers in the figures do not correspond to the retinal anatomy. While a realistic thickness of 90 μm for controls was presented for the routine assessment of the peripapillary retinal nerve fiber layer (Fig. 2), neither the results section nor the tables reported values within this expected range. Instead, unitless data were presented with unrealistic values for single layer thickness ranging from 4.19 to 7.81, with a supposed total retinal thickness of 62.6 in controls. Published data on paramacular retinal thickness from control subjects are around 320 μm and around 8 mm3 for retinal volume.2–4 More confusingly, the reported thickness values of the different layers relative to each other did not correspond to retinal anatomy. For example, the highest values were reported for the retinal pigment epithelium, which is impossible because anatomically, this is one of the thinnest retinal layers. Furthermore, how do the authors propose to be able to measure the thickness of a cellular monolayer, the inner limiting membrane, with values exceeding most of the other retinal layers?  
The segmentation software's results cannot be compared with histologic studies because OCT makes a determination of the retinal layers based on changes in color and image saturation and thus the analysis is very different from that of retinal anatomy studies using stains and other histologic techniques. 
The 10 segmentation lines generated by the automated segmentation software indicate the borders between distinct retinal layers and the software provides automated measurements of the layers based on a determination of the limits by the prototype. The prototype delimits these borders in each image and provides thicknesses of 10 retinal layers in a spreadsheet database (Excel; Microsoft Corp., Redmond, WA, USA) that the software generates automatically. 
The measurements of the layer thicknesses are accurate and no mistakes are possible because the software exports these automated measurements. 
We think that the layer thickness measurements are determined by the prototype as the distance between two consecutives layers (e.g., the retinal pigment epithelium thickness may be calculated as the distance between the lines assigned by the prototype as the RPE/Bruch's complex and the outer photoreceptor segments). In our opinion, the measurements of the prototype should be used for comparisons with healthy subjects or to look for models of retinal layer atrophy in different diseases, but not to compare with histologic studies. 
Histologic analysis present variations in the thickness of different retinal layers depending on the tissue processing technique (e.g., layers with higher eosin absorption may appear to have a greater thickness). Both histology and OCT are interpretative techniques, but neither represents 100% reality. In our opinion, the main strength of segmentation analysis in systemic diseases is to analyze which layers are more affected. The number of microns of each layer, as measured by a particular technique, has no clinical application. 
In addition to these methodological errors, the authors incorrectly labelled the retinal layers in Figure 3, where the nerve fiber layer is indicated while the segmentation lines encompass the ganglion cell layer and the ganglion cell layer is indicated while the lines segment the inner plexiform layer
Upon review, the article authors detected a mistake in Figure 3 that requires correction: the name of the nerve fiber layer should be “ganglion cell layer” and the name of the ganglion cell layer should be “inner plexiform layer.” We have now modified this figure and submitted the revised figure. 
There are exciting times ahead for the application of retinal OCT in neurodegenerative disease5 and a rigorous, quality controlled approach (e.g., using the OSCAR-IB criteria) will be needed to establish the technique as a potential outcome measure for clinical trials.6  
We agree with the importance of clarifying the figure errors. Thank you for your correction and for improving our paper. 
Albrecht P Balk L Oberwahrenbrock T Petzold A Paul F. Retinal Segmentation to demonstrate hyperplasia in ataxia of Charlevoix-Saguenay: critique on study methodology and results. Invest Ophthalmol Vis Sci . 2014; 55: 4728. [CrossRef] [PubMed]
Garcia-Martin E Pablo LE Gazulla J Retinal segmentation as noninvasive technique to demonstrate hyperplasia in ataxia of Charlevoix-Saguenay. Invest Ophthalmol Vis Sci . 2013; 54: 7137–7142. [CrossRef] [PubMed]

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.