Purchase this article with an account.
Andreas Wahle, Adam T Harding, Kyungmoo Lee, Mona K Garvin, Wallace L M Alward, John H Fingert, Teresa R Kopel, Michael David Abramoff, Young H Kwon, Milan Sonka; Supporting Glaucoma Structure/Function Research with a Unified OCT, Fundus, and Visual Field Database. Invest. Ophthalmol. Vis. Sci. 2014;55(13):4821.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The relationship between structural changes and visual function in glaucoma patients is of major interest. Engineering methods provide quantitative tools to evaluate, e.g., cell-layer thickness and visual-field distribution. Ophthalmic data are highly heterogeneous, located on various scanners, and frequently not standardized. To reduce the burden on researchers, we developed a highly automated data-submission and storage system based on the open-source NIH-funded eXtensible Neuroimaging Archiving Tool (XNAT, http://www.xnat.org/). We extended XNAT with ophthalmology-specific definitions (included in XNAT 1.5) and custom upload scripts, thus providing a centralized location for the data collected with automatic translation into standardized formats like DICOM.
Prospective data from 132 glaucoma patients of varying severity was collected: OCT, Fundus, Visual Field (VF). We developed software to automatically anonymize, catalog, and associate those disparate datasets, translate proprietary OCT formats and Fundus photos into DICOM, and combine report data in PDF files. After collection from the scanners, datasets are uploaded without supervision in batches of mixed data types using drag-and-drop operation (Fig. 1). For easier image processing of large data volumes, the system also provides a programmatic interface; complex tasks such as retinal OCT layer segmentation are performed without manual interaction.
The current dataset contains 1482 Heidelberg Spectralis OCT (mostly 9-field) and 525 Zeiss Cirrus OCT with associated report pages, fundus photographs, and Humphrey VF reports, from up to 3 visits. All datasets were successfully uploaded to the XNAT system, up to several hundred scans per batch. Some scans (<1%) had to be re-uploaded due to network issues or typos in the anonymization information. Both the anonymized original and converted/standardized versions of the scans are available, thus equally allowing processing with vendor-provided or self-developed tools.
Accessing and converting data directly from the scanner to perform research is difficult and time consuming. We developed and deployed a customized system for centralized storage of related structural and functional data. This will allow efficient analysis of these relationships in glaucoma and other research without the need to handle various data formats and access points.
This PDF is available to Subscribers Only