Purpose
To evaluate a texture analysis method to assess the integrity of the photoreceptor layer (PRL) in spectral domain - optical coherence tomography (SD-OCT) images of patients suffering from inherited and acquired macular diseases.
Methods
High-quality SD-OCT images were acquired over a 10°x15° retinal area centered on the fovea in four patients suffering from macular diseases, two of which had a previous diagnosis of age-related macular degeneration and the others of rod-cone dystrophy. Eight age-matched healthy subjects were used as controls. After automated segmentation of the PRL, a sliding box algorithm was used to calculate the Gray Level Co-occurrence Matrix (GLCM) over the PRL. From the GLCM matrix, three texture parameters were computed, which included the contrast, the local homogeneity, and the correlation (256 gray levels; directions 0°, 45°, 90° and 135°). The raw data were used to generate three statistical descriptors of the PRL integrity, such as the ratio of contrast, the ratio of homogeneity and the ratio of correlation; all the parameters were calculated as the ratio between 0° and 90° directions. Texture analysis of the PRL was then performed over the foveola (0.25 mm diameter) and other regions of interests (ROIs) across the temporal and nasal meridians, such as the parafovea (0.26-0.86 mm), the perifovea (0.87-2.00 mm) and the mid-periphery (2.01-2.36 mm).
Results
GLCM analysis was valuable to characterize the structure and texture of the regular horizontal stratification of the PRL in SD-OCT images. The method was highly sensitive for assessing the pathological changes of the ellipsoid zone in patients compared with age-matched controls. In patients, all the statistical descriptors based on texture analysis of the PRL were far below the 95% confidence interval of controls in all ROIs, except for the mid-periphery.
Conclusions
The GLCM based imaging biomarkers showed to be valuable to assess quantitatively the pathological disruption of the PRL. Texture analysis of the PRL in high-resolution SD-OCT images of the retina represents a valuable objective tool for monitoring and assessing the integrity of photoreceptors in patients.