Purchase this article with an account.
Alfredo Ruggeri, Pedro Guimarães; Automatic estimation of short- and long-range tortuosity of corneal sub-basal nerves. Invest. Ophthalmol. Vis. Sci. 2016;57(12):1923.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To automatically grade corneal nerve images according to two classes of nerve tortuosity (NT), representing respectively short-range or long-range directional changes.
The clinical perception of corneal NT was recently shown to have two distinct forms, characterized by either short-range (high frequency, low amplitude) or long-range (low frequency, high amplitude) directional changes (Lagali et al., IOVS Aug 2015). Ground-truth (GT) for each NT form in 30 confocal microscopy images was provided by 7 expert graders, who visually examined the images and ordered them by increasing nerve tortuosity, according to either tortuosity definition (see Fig.1 for examples). Their average orderings were assumed as GT.Using the same 30 images, we traced corneal nerves with a custom computerized procedure and computed for each image more than one hundred NT metrics, using several different mathematical formulations, including angle-, twist- and curve-based metrics. The ranking estimated by each formulation was then compared to the GT (for both short- as well as long-range NT) using the Spearman Rank Correlation (SRC) coefficient.
Table 1 presents the SRC coefficients for the best-performing NT metrics. It appears that angle-based metrics are the best to capture short-range NT, whereas twist-based metrics are the best for expressing long-range NT.
The results indicate which of the many mathematical formulations for NT we considered is the best for describing either the short- or the long-range NT. The next step will be to assess to what degree the various corneal pathologies exhibit one or the other NT class in their sub-basal nerve patterns.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.
Figure 1: Corneal nerves image with the highest tortuosity rank for the short-range (top) and the long-range tortuosity (bottom).
Table 1. SRC coefficients with respect to the short-range GT or the long-range GT for angle-based (top), curve-based (middle) and twist-based (bottom) NT metrics.
This PDF is available to Subscribers Only