pawFLIM: reducing bias and uncertainty to enable lower photon count in FLIM experiments

Abstract

Förster resonant energy transfer measured by fluorescence lifetime imaging microscopy (FRET-FLIM) is the method of choice for monitoring the spatio-temporal dynamics of protein interactions in living cells. To obtain an accurate estimate of the molecular fraction of interacting proteins requires a large number of photons, which usually precludes the observation of a fast process, particularly with time correlated single photon counting (TCSPC) based FLIM. In this work, we propose a novel method named pawFLIM (phasor analysis via wavelets) that allows the denoising of FLIM datasets by adaptively and selectively adjusting the desired compromise between spatial and molecular resolution. The method operates by applying a weighted translational-invariant Haar-wavelet transform denoising algorithm to phasor images. This results in significantly less bias and mean square error than other existing methods. We also present a new lifetime estimator (named normal lifetime) with a smaller mean squared error and overall bias as compared to frequency domain phase and modulation lifetimes. Overall, we present an approach that will enable the observation of the dynamics of biological processes at the molecular level with better temporal and spatial resolution.