Alejandra Consejo1, D Robert Iskander2. 1. Department of Biomedical Engineering, Wroclaw University of Science and Technology, Poland. Electronic address: alejandra.consejo@pwr.edu.pl. 2. Department of Biomedical Engineering, Wroclaw University of Science and Technology, Poland.
Abstract
PURPOSE: To accurately and precisely demarcate the transition points in the anterior limbal surface from 3D anterior eye height data. METHODS: Data of anterior eye surface was acquired using a corneo-scleral profilometer. Two methods for automatic limbal radius estimation were proposed. One is based on the residual error between the original data and low order fit data while the other method takes a step further, calculating the cumulative root mean square (RMS) of the residual error. Data of a simulated example, collected from an artificial bi-sphere test surface and four real eyes is used to validate the methodology. RESULTS: Both methods of limbal radius estimation were found to be equivalent in a low noise regime while the residual RMS method performs better in higher noise regimes. The relative error for the proposed methods was below 0.05% and below 1.1% for the simulated test surface and the artificial bi-sphere, respectively. For the four considered subjects, the variability of the methods was less than 10μm. There was no statistically significant difference between methods nor between method and subject but there was statistically significant difference between the subjects (two-way ANOVA, p<0.001). Asymmetry in limbal shape between nasal and temporal side was observed. CONCLUSIONS: Determining the anterior limbus position in a non-invasive, automatic, accurate and precise manner is possible using 3D anterior eye height data. The topographical information of limbus does not necessarily coincide with the maximum rate of change in colour from the iris to sclera observable with en-face imaging.
PURPOSE: To accurately and precisely demarcate the transition points in the anterior limbal surface from 3D anterior eye height data. METHODS: Data of anterior eye surface was acquired using a corneo-scleral profilometer. Two methods for automatic limbal radius estimation were proposed. One is based on the residual error between the original data and low order fit data while the other method takes a step further, calculating the cumulative root mean square (RMS) of the residual error. Data of a simulated example, collected from an artificial bi-sphere test surface and four real eyes is used to validate the methodology. RESULTS: Both methods of limbal radius estimation were found to be equivalent in a low noise regime while the residual RMS method performs better in higher noise regimes. The relative error for the proposed methods was below 0.05% and below 1.1% for the simulated test surface and the artificial bi-sphere, respectively. For the four considered subjects, the variability of the methods was less than 10μm. There was no statistically significant difference between methods nor between method and subject but there was statistically significant difference between the subjects (two-way ANOVA, p<0.001). Asymmetry in limbal shape between nasal and temporal side was observed. CONCLUSIONS: Determining the anterior limbus position in a non-invasive, automatic, accurate and precise manner is possible using 3D anterior eye height data. The topographical information of limbus does not necessarily coincide with the maximum rate of change in colour from the iris to sclera observable with en-face imaging.
Authors: Ahmed Abass; Bernardo T Lopes; Ashkan Eliasy; Richard Wu; Steve Jones; John Clamp; Renato Ambrósio; Ahmed Elsheikh Journal: PLoS One Date: 2018-11-26 Impact factor: 3.240