Purchase this article with an account.
Zohreh Hosseinaee, Le Han, Olivera Kralj, Alexander Wong, Luigina Sorbara, Kostadinka K Bizheva; Fully automated segmentation algorithm for corneal nerves analysis from in-vivo UHR-OCT images. Invest. Ophthalmol. Vis. Sci. 2019;60(9):167.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To develop a fully automated algorithm for segmentation and morphometric analysis of corneal nerves from UHR-OCT images acquired in-vivo from human corneas.
Volumetric (700 x 700 x 2048) images of the anterior human cornea were acquired from ~1mm x 1mm area in-vivo with a 250 kHz SD-OCT system that provided ~1.5 μm isotropic resolution in corneal tissue. Images of the sub-basal corneal nerves were processed with a novel segmentation algorithm that includes multiple steps: 1. edges and pixel connectivity enhancement; 2. background subtraction, enhancement of tubular structures, and segmenting the nerves; 3. morphological operations for further enhancement of the detected nerves. Corneal nerves morphometric parameters such as length, number of branching points and tortuosity were determined from the segmentation data. Nerves were segmented manually too, for direct comparison with the results from the automatic segmentation.
Fig.1A shows a representative maximum projection en-face UHR-OCT image of the sub-basal corneal nerves in a healthy human subject. Fig. 1B and C show the output image of the first and second step respectively. Fig 1.D represents the output for the proposed automatic corneal nerve segmentation and Fig 1.E, shows the manual nerve tracing output. Comparison between manually and automatically segmented corneal nerves showed excellent correlation.
The fully automatic corneal nerve segmentation algorithm proposed in this study can aid the morphometric characterization of corneal nerves and can be applied to the diagnostics and monitoring the treatment of corneal diseases such as diabetic peripheral neuropathy.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.
Figure 1. Automatic corneal nerve segmentation system, (A) Original corneal image. (B) Preprocessing stage image output. (C) Morphological stage image output. (D) Automatically segmented corneal nerves overlaid on the original image. (E) Manually traced corneal nerves.
This PDF is available to Subscribers Only