Purchase this article with an account.
B. L. Petrig, A. E. Elsner, D. A. VanNasdale, B. P. Haggerty, Y. Zhao, S. A. Burns; Detection of Retinal Disease by Cross-Correlation of Different Imaging Modes. Invest. Ophthalmol. Vis. Sci. 2008;49(13):5395. doi: https://doi.org/.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To test the hypothesis that in normal eyes, the well-ordered retinal layers are relatively transparent and the light return is mainly specular, while in retinal disease, the tissues become disordered and pathology leads to increased scattered light. This implies that images of the deeper layers are less well correlated with those of superficial layers in diseases such as age-related macular degeneration (AMD) compared to normal eyes. We compared images obtained with different imaging modes, e.g. specular light vs. multiply scattered light, which emphasize superficial vs. deep retinal layers using cross-correlation techniques.
Sixteen patients, classified by clinical criteria to have early AMD, mean age 69 years, and 28 controls, mean age 68 years, were imaged using confocal scanning laser ophthalmoscopy with near infrared illumination. The grayscale values of directly backscattered light images and multiply scattered light images were compared by a zero-lag cross-correlation algorithm. When derived from the same raw image data, both computed image types are inherently aligned, and the zero-lag cross-correlation coefficient was used. With two sequential data sets, images were first aligned using one or more landmarks. The cross-correlation function was plotted in 3-D to assess alignment accuracy and strength of correlation using peak value and peak width. The distribution, mean and sources of variability of the cross-correlation were examined, with the means compared by ANOVA.
Multiply scattered light images appeared strikingly different from directly backscattered light images for patients with AMD, but sometimes confined to focal changes. The zero-lag cross-correlation coefficients of 14 deg regions overlapped for AMD patients and controls, 0.54 to -0.02 and 0.72 to 0.12, respectively. The mean coefficient was smaller, but not significantly so, for AMD vs. controls (0.35±0.13 vs. 0.43±0.18, p<0.15). The multiply scattered light images frequently enhanced drusen and pigmentary changes in controls, potentially decreasing correlation coefficients.
Multiply scattered light images were never well-correlated with directly backscattered light images in our older subjects. Drusen and pigmentary changes may lead to poor correlation between directly backscattered and multiply scattered light images.
This PDF is available to Subscribers Only