Premium
The distribution of retinal thickness in healthy eyes and its use in the objective analysis of optical coherence tomography (OCT) scans
Author(s) -
HALDAR S,
DAVIES N
Publication year - 2011
Publication title -
acta ophthalmologica
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.534
H-Index - 87
eISSN - 1755-3768
pISSN - 1755-375X
DOI - 10.1111/j.1755-3768.2011.2216.x
Subject(s) - optical coherence tomography , kurtosis , skewness , standard deviation , retinal , medicine , ophthalmology , fundus (uterus) , nuclear medicine , mathematics , statistics
Purpose OCT provides quantitative data but interpretation remains subjective. We have developed a statistical method to detect abnormal retinal thickness in OCT scans. Methods High resolution macula topography maps, which were identified as normal by a retinal specialist, were exported from a spectral domain OCT/SLO machine. A Matlab® algorithm was written to perform image registration to a model macula. A reference map was constructed for each eye consisting of mean thickness and standard deviation at each point. Kurtosis and skewness were calculated. To analyse pathological scans areas > 2 SD from the mean were deemed abnormal. Results 151 left eye scans and 112 right eye scans were analysed. The mean foveal thickness was 190.7 microns (SD 1.8) for the left and 189.2 microns (SD 1.6) for the right. Skewness and kurtosis were assessed over the macula. The mean skewness was 0.34 (‐0.39 to 0.39) for the left and ‐0.0154 (‐0.45 to +0.45) for the right. The mean kurtosis was 0.95 (SE 0.39, range ‐0.8 to 0.8) for the left and ‐0.0324 (SE 0.45, range ‐0.90 to 0.90) for the right. 20 abnormal scans from patients with diabetes were aligned with the reference maps. The mean percentage area of abnormality >2 SD was 28%, 2‐3 SD was 8%, 3‐5 SD was 7%, 5‐10 SD was 8% and >10 SD 5%. There was a linear correlation of area >2 SD with areas 3‐5 SD (R2 = 0.65) and 5‐10 SD (R2 = 0.73) indicating that larger lesion area is associated with greater oedema. Conclusion The data show a normal distribution of retinal thickness in healthy eyes. A quantified image of statistical abnormality in a diseased eye can be generated which is helpful for standardising interpretation.