Better estimates of exponential decay parameters

Video-rate fluorescence lifetime imaging camera with CMOS single-photon avalanche diode arrays and high-speed imaging algorithm

“…The data was stored in memory and then post-processed using maximum-likelihoodestimation (MLE) or least-square method (LSM)-based curvefitting software pixel-by-pixel to generate a lifetime image. Although the MLE (or LSM) has merits of wide resolvability range and high photon efficiency, and usually hundreds of photons are enough to reach an acceptable accuracy, 14,15 the data acquisition is still slow (measurement time = N x × N y /f p , where f p is the scanning frequency and N x × N y is the dimension of the array) and limits the systems to imaging only stationary objects. In many biological applications, real-time FLIM imaging for monitoring cell dynamics in low light level is desirable.…”

Bayesian analysis of fluorescence lifetime imaging data

“…1 When significant counts are available, where Gaussian statistics closely approximate the Poissonian statistics associated with photon count data, data fitting is relatively straightforward and both least squares (LS) and maximum likelihood estimation (MLE) have been shown to perform well. [2][3][4][5][6][7] The collection of significant counts is however inevitably linked to longer experimental acquisition times. When only a low number of counts is available, resulting from either low fluorophore concentrations, or their propensity to fade or when dynamic processes need to be studied, difficulties in determining the correct lifetime, or establishing the error associated with lifetime estimation inevitably increase.…”

“…When only a low number of counts is available, resulting from either low fluorophore concentrations, or their propensity to fade or when dynamic processes need to be studied, difficulties in determining the correct lifetime, or establishing the error associated with lifetime estimation inevitably increase. Under such conditions, standard LS has been shown to yield poor estimates, 2,3,5,7 though data manipulation prior to analysis 4,5 and a modification of the LS weighting 8 have both been shown to improve LS estimates. The use of MLE methods for the analysis of exponential decays has also been thoroughly studied 3,5 and although it does yield parameter predictions that are more accurate than those of LS 3 and offer good resilience to changes in bin width and photon count, 5 the critical dependence of MLE algorithms on their starting location (in the parameter space) lead to the suggestion of a hybrid LS-MLE Further author information: Mark Rowley: E-mail: mark.rowley@kcl.ac.uk; Paul Barber: E-mail: Paul.Barber@rob.ox.ac.uk analysis.…”

“…2 Though MLE and LS do perform optimally for large sample sizes, 9 it has been suggested that when the large sample assumptions are not met the Bayesian framework should be a preferred alternative analysis technique. 9 The effectiveness of such FLIM techniques to study the cellular environment is, therefore, critically dependent on the accuracy of the data analysis methods.…”

Detectors for single-molecule fluorescence imaging and spectroscopy

“…Simultaneously, the nanotime information can be used to build fluorescence decay histograms for each pixel in each frame, or for each identified particle (or region of interest) in each frame. Standard fluorescence lifetime fitting procedures can be used (for instance, maximum likelihood estimation -MLE -of single lifetimes (150), or non-linear least-square fitting of multiple lifetime or stretched exponentials) to extract one or more numerical value per pixel. In the simple case where a single lifetime is extracted, this information can be color-coded and represented as a new, fluorescence lifetime image (in addition to the intensity image obtained initially) or a FRET efficiency image if the observed signal is that of a donor molecule in a FRET pair.…”