Purchase this article with an account.
M. Klemm, A. Dietzel, E. Nagel, D. Schweitzer, U. Graichen, J. Haueisen; Statistical Analysis of Reproducibility of Fluorescence Decay Histograms at the Fundus. Invest. Ophthalmol. Vis. Sci. 2010;51(13):2293.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To determine the reproducibility of fluorescence lifetime imaging (FLIM) measurements in ophthalmology on the basis of an evaluation of fluorescence decay photon histograms.
To measure the fluorescence lifetime at the human fundus a modified Heidelberg Retina Angiograph was used. The endogenous fluorophores were excited by a diode laser with pico-second pulses at 446 nm and a repetition rate of 80 MHz. The auto-fluorescence was detected in two spectral channels: 490-560 nm and 560-700 nm using the time-correlated single photon counting method. A time resolution of approximately 12.2 ps was achieved by dividing the time between laser pulses (12.5 ns) into 1024 time channels. The acquired images cover 30° of the fundus with a lateral resolution of 40 x 40 µm². For this study six healthy male subjects aged between 25 and 35 years were examined. Each subject was measured twice per day (with one hour break in-between) on two days a week over three weeks resulting in a total of ten datasets per subject. Three regions (fovea centralis; a blood vessel free region of the papilla-macula-bundle; optic disc) each with a size of 15 x 15 pixels were manually selected and binned into a single fluorescence decay photon histogram. In those histograms the offset (background and detector dark current) was subtracted, the pre-stimulus part was removed and the maxima were normalized. A two-sample Kolmogorov-Smirnov (KS) test was performed for all unique permutations of the ten repetitive measurements separately for each spectral channel at a 5 % significance level.
The KS test showed that the measured photon distributions were significantly different in the short wavelength channel for about 6 % and in the long wavelength channel for about 8 % over all regions and subjects. The largest differences occurred in the optic disc region (12 % on average) due to the generally low photon counts while the smallest differences occurred in the macula region (1 % on average). Comparing both spectral channels with the KS test stated in about 95 % of the cases different distributions proving that there is different information in both channels. The comparison of the first and the second measurement per day did not reveal significant differences.
The reproducibility of this new method for functional diagnostics of metabolic changes allows for the reliable application of subsequent analysis and diagnosis methods.
This PDF is available to Subscribers Only