COMPSTAT 2016: Start Registration
View Submission - CRoNoS FDA 2016
A0176
Title: A functional regression approach for modelling the retinal nerve fiber layer thickness in the eyes of healthy subjects Authors:  Eleonore Pablik - Medical University of Vienna (Austria) [presenting]
Florian Frommlet - Medical University Vienna (Austria)
Abstract: One hundred healthy volunteers underwent ophthalmic examination where the retinal nerve fiber layer (RNFL) thickness was measured in a circle of 3.4 mm diameter around the optic disc with Fourier-domain optical coherence tomography. Measurements were obtained in 256 equidistant sectors on the circle. Functional data analysis was used to develop a model to partly explain the inter-subject variability of RNFL in these healthy subjects in order to obtain a narrower range of normative RNFL data. The long term goal of this modelling is to improve the diagnosis of early glaucoma or other diseases affecting the RNFL thickness. As a first step we used landmarking to provide adjusted percentiles for each of the 256 points of measurement instead of the raw data's percentiles. In the second step we modelled the patient-specific variation around the adjusted median with 16 basis functions. Due to the fact that a positive correlation between retinal nerve layer thickness and the amount of retinal vessels could already be shown in earlier publications we tried to explain the magnitude of each of these basis functions with the amount of retinal vessels in the corresponding area using linear regression models. We compare our functional regression approach with a standard multiple regression approach previously used in the literature.