Project: The Relation between Retinal and Optic Nerve Head Parameters and Circumpapillary Retinal Nerve Fiber Layer Profile
Funding: Research Grant - Vienna Science and Technology Fund (Life Sciences Call 2011)
Team members at MIM: Georg Fischer (project leader), Ivania Pereira, Florian Schwarzhans
Partners: Clemens Vass (Department of Ophthalmology and Optometry, Medical University Vienna)
Background: Measurement of retinal Nerve Fiber Layer (RNFL) is of major importance for early glaucoma diagnosis. However, its high variability in the normal population may lead to a wrong diagnosis. In order to determine an independent set of retinal and optic nerve head parameters that may be related with the distribution of RNFL thickness (RNFL profile), an automatic method of image processing will be developed. A multivariate model describing the relation between the retinal parameters and the RNFL profile will be developed and validated in an independent sample based on the output of the image processing measurement techniques. In the future, the validated model opens the perspective of compensation for the intersubject variability of RNFL measurements. This will benefit the early diagnosis of Glaucoma.
Methods: Retinal Nerve Fiber Layer (RNFL) is a retinal layer composed by the extension of cell axons of the retinal ganglion cells and has been a relevant concern in ophthalmic research in the past years, due to its diagnostic value for early glaucoma diagnosis. When measuring its thickness in the vicinity of the optic disc in healthy subjects, one can observe a large variability of values, which decreases sensitivity and specificity of RNFL measurements and its diagnostic value, leading sometimes to false conclusions in diagnosis. It is, then, of major relevance for clinical concerns, to understand the reasons of this intersubject variability and which factors may correlate with cell axon distribution. The relation between the major retinal vessel location and RNFL maxima location and the circumpapillary RNFL distribution (RNFL profile) has been proven, as well as the relation between optic disc size and RNFL thickness. It is now important to investigate whether other parameters may be correlated with the RNFL profile and whether this relation can be modeled in a multivariate way. We focus on parameters that are independent from the imaging device, but are representative of the physiologic characteristic of the population and present themselves a high variability in healthy population. These parameters comprise a circumpapillary distribution density function of retinal vessels, descriptors of size and shape of the optic disc and of the topographic relation between the fovea and the optic disc.
In order to obtain the referred parameters, algorithms for automatic feature detection are developed using intensity images from High Definition Optical Coherence Tomography from a sample of 120 healthy subjects. This process integrates image normalization and enhancement steps, based on contrast and luminance drifts, vessel extraction based on active contours theory and skeleton decomposition based on region growing and morphological operators. RNFL data will be parameterized and the obtained coefficients will be correlated with the retina and optic disc parameters. The significant parameters are selected for multiple regression analysis, in order to determine a model of the relation between the retinal and optic disc parameters and the parameterized RNFL profile. The model will then be validated in an independent sample of at least 120 healthy subjects.