Purchase this article with an account.
Neha Anegondi, Simon S. Gao, Jasmine Patil, Alexandre Fernandez Coimbra; Evaluation of shape-descriptive and texture features as potential prognostic variables in progression of geographic atrophy.. Invest. Ophthalmol. Vis. Sci. 2019;60(9):1906.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To assess the prognostic potential of multiple shape-descriptive (Pfau, M., et al. Retina. 2018) and textural features in the progression of geographic atrophy (GA).
A retrospective analysis was performed using the monthly spectral domain optical coherence tomography (SD-OCT) image data from patients in the MAHALO (NCT01229215) study. OCT images with significant motion artifacts were excluded. Thirty-three eyes [14 fellow eyes (FE) and 19 sham study eyes (SE)] were evaluated for this analysis. The GA lesion areas were derived from en face choroidal images at baseline and 18-month using a semi-automated algorithm (Figure 1A to 1D). Shape-descriptive [lesion area (mm2), perimeter (mm), circularity, feretmax (mm), feretmin (mm), square-root of lesion area, perimeter and circularity] and textural (contrast, correlation, energy and homogeneity) parameters were calculated based on the algorithm output and correlated with the change in GA lesion area using simple and multiple linear regression models. Coefficient of determination (R2) values were reported.
Linear regression of each shape-descriptive parameter with lesion growth (determined by taking absolute difference between baseline and 18-month lesion areas) showed high R2 values of 0.28, 0.24 and 0.23 for square-root area, feretmin and square-root circularity, respectively (Table 1). Similar analysis with textural parameters showed energy to have the highest R2 value of 0.1 (Table 1). Multiple regression of all shape-descriptive parameters with lesion growth showed R2 value of 0.36. Similarly, all textural parameters showed R2 value of 0.14.
These findings suggest that shape-descriptive and textural factors may have potential as prognostic variables for GA progression. Combinations of shape-descriptive and textural factors will need to be further explored in larger datasets to evaluate their validity in predicting GA lesion growth.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.
Table 1: Mean (standard error of mean) of shape-descriptive and textural parameters at baseline and 18-month, respectively. R2 is the coefficient of determination obtained from linear regression between baseline value of each parameter and lesion growth (absolute difference between baseline and 18-month lesion areas).Sqrt., square-root
Figure 1: En face choroidal images from a patient at (A) Baseline and (B) 18-month with corresponding GA segmentation (C-D).
This PDF is available to Subscribers Only