Purchase this article with an account.
Anna Dastiridou, Akram Belghith, Linda M Zangwill, Robert N Weinreb, Alex S Huang; Three-Dimensional 360 Degrees Imaging of Aqueous Humor Outflow Structures in the Living Human Eye with Spectral-Domain OCT. Invest. Ophthalmol. Vis. Sci. 2016;57(12):5119.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To create a three-dimensional model of the circumferential aqueous humor outflow (AHO) structures in the living human eye using an automated detection algorithm of Schlemm’s Canal (SC) and first-order collector channels (CC) applied to non-invasive spectral domain optical coherence tomography (SD-OCT) images.
High-resolution SD-OCT images from volume datasets of angle structures from a living human eye (34 year-old healthy volunteer) were acquired (Heidelberg Engineering; Spectralis) with the anterior segment module (scleral mode; ART=9; resolution axial/lateral/B-to-B; 3.9/11/11 μm). Overlapping volume scans were manually set circumferentially around the limbus. A Bayesian Ridge method was used to approximate the location of the SC on the infrared confocal laser scanning ophthalmoscopic (CSLO) images with a Cross Multiplication tool developed to initiate SC and CC detection automated through a Fuzzy Hidden Markov Chain approach. Individual B-scans were organized 360 degrees around the limbus anchored to the CSLO image. The guiding principal of the automated detection method was to set error tolerance such that missing structures (false negatives) would be prioritized over creating false structures (false positives).Automatic segmentation of SC and first-order AHO pathways were manually confirmed by two masked graders. The following parameters were graded: complete false negative detection of SC, complete false positive detection of SC, partial detection of SC (<50% of the true SC), exaggerated detection of SC (>200% of the true SC), complete false negative and complete false positive detection of CC.
48 out of 5114 (<1%) scans were deemed ungradable. Overall, the automatic segmentation algorithm performed well, with 1.5% out of 5066 images showing false negative SC detection, 0.7% false positive SC detection, 3.8% partial SC detection, 0.1% exaggerated SC detection, 29.5% false negative CC detection and 1.2% false positive CC detection. The agreement between the two graders in each parameter tested was good (kappa ranging between 0.63-0.78), with the exception of exaggerated SC detection (kappa=0.2), where the incidence was extremely low.
360-degree imaging of AHO structures in the living human eye is possible and can provide information about the outflow pathways.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.
This PDF is available to Subscribers Only