Purchase this article with an account.
Andreas Wahle, Kristine E Lee, Kyungmoo Lee, Li Zhang, Hrvoje Bogunovic, Adam T Harding, Todd E Scheetz, Milan Sonka, Ronald Klein, Michael David Abramoff; Use of XNAT to generate phenotype data from OCT scans for use with genetic association analyses. Invest. Ophthalmol. Vis. Sci. 2015;56(7 ):1261.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The correlation of phenotypic indices like thickness of layers or pathological conditions with genetic markers is of high interest. However, image data, analyses from manual or automated segmentation, and genetic information are usually maintained in separate systems with proprietary data formats and are difficult to integrate. We use the NIH-supported XNAT system [http://www.xnat.org/] to convert and centrally store OCT data and to apply image-processing techniques to obtain phenotypic indices. A custom programmatic interface is used to automatically extract phenotypic indices and to produce the necessary input for PLINK, an open-source whole genome association analysis toolset.
OCT data are converted to DICOM format and uploaded to XNAT as a batch using custom scripts. Layer segmentation is fully automated, followed by segmentation of choroidal vasculature and detection of fluid-filled abnormalities, controlled by pipeline processing. Phenotypic indices can be tailored for given studies and automatically calculated from the segmentation results from all or a subset of patients as arrays of scalar values. Those are converted into PLINK phenotype files and fed along with the participants' genotype information into the PLINK tool for statistical evaluation of genotype-phenotype associations.
As a proof of concept, 7,386 Topcon SD-OCT scans from 1,879 participants from the BEAVER Dam Eye Study (BDES) and Wisconsin Epidemiologic Study of Diabetic Retinopathy (WESDR) projects performed at the University of Wisconsin were uploaded to an XNAT 1.5 instance at the University of Iowa. Segmentations were performed for 11 layer boundaries, then choroidal vasculature and fluid-filled abnormalities detected, all stored with the scans in XNAT. Sets of phenotypic indices such as layer thicknesses or computed scalar values from multiple layers were selected, resulting in the automated generation of the phenotype input for PLINK in the expected format.
The combination of the original image data with the results of segmentations yielding phenotypic indices offers unique opportunities to perform phenotype/genotype studies. The XNAT system and its programmatic REST interface along with custom scripting to interface with the popular PLINK genome association tool have proven to be valuable in efficiently performing this process with minimal user interaction.
This PDF is available to Subscribers Only