Abstract
Purpose :
To develop an algorithm for automated quantification of Haller's layer in choroid using swept-source optical coherence tomography.
Methods :
The primary contribution of this work involves defining the approach
for detecting the boundaries of Haller's and Sattler's layer. In brief, the proposed
algorithm extracts the choroidal vessel cross-sections using novel exponentiation-based
binarization. Subsequently, it detects the large choroidal vessels based on statistically
defined median criteria. Finally, the desired boundary is obtained by extrapolating and
smoothening the innermost points of the large vessel cross-sections. On 50 OCT B-scans
of 50 healthy subjects, algorithm is validated, both qualitatively and quantitatively,
vis-a-vis intra-observer variability. A thorough statistical analysis has been performed
using various metrics including Dice coefficient (DC), correlation coefficient (CC) and
absolute difference (AD).
Results :
The proposed algorithm achieves a mean DC of 89.48% (SD:5.03%) which is
in close agreement with corresponding intra-observer repeatability value of 89.12%
(SD:5.68%). Similarly, proposed algorithm achieves mean AD and mean CC of 17.54 m
(SD:16.45 m) and 98.10% (SD:1.60%) which are close to corresponding intra-observer
repeatability values of 19.19 m (SD:17.69 m) and 98.58% (SD:1.12%), respectively.
Conclusions :
Our study demonstrates high correlation between algorithmic and
manual delineations and can be deployed for further clinical applications to analyze
choroid in greater depth, especially in diseased eyes.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.