Abstract
Purpose:
Quantitative autofluorescence (qAF) provides images in which the grey levels are calibrated against a fluorescent standard and are corrected for ocular media absorption and refraction (magnification). Because the AF is not uniformly distributed at the posterior pole, previously, we have only used average qAF in large areas for clinical studies. To remedy this limitation, we developed an image analysis method to rapidly assess whether localized areas have normal or abnormally high or low autofluorescence levels.
Methods:
The qAF image of any patient was compared point by point to the spatial qAF distribution of healthy eyes of the same age and race. These distributions were derived from data for both eyes in 97 healthy subjects with normal retinal status (Greenberg et al. IOVS. 2013). The images were spatially normalized to the distance between the optic disk edge and the fovea. Mean and 95% confidence intervals of qAF in 29 specific retinal areas were calculated for different races, ages and eyes. 8th-order polynomials were fitted to these data providing a model of the qAF distribution for a field of about 20 degrees in diameter (centered on the fovea) for a given eye, race, and age. The patient image was then compared at each location to the corresponding distribution using Z-score statistics.
Results:
Color-coded Z-score maps in several retinal pathologies show areas where the AF is significantly higher (Z>2, p<0.05) or lower (Z<-2) than the 95% confidence interval.
Conclusions:
qAF Z-score maps may be particularly useful for assessing pathologies characterized by subtle localized changes in AF, such as in RP and AMD, and for detecting significant early changes in retinal dystrophies.