Section: New Results

Lifetime estimation in photon counting-based fluorescence lifetime imaging microscopy

Participants : Philippe Roudot, Charles Kervrann.

In this study, we investigated a Maximum Likelihood (ML) framework for photon counting-based fluorescence lifetime estimation in Fluorescence Lifetime Imaging Microscopy (FLIM). Data collected at a given pixel consist of photon counts exponentially decreasing along the time and are assumed to follow Poisson statistics (see Fig. 6 ). A careful analysis of the biophysical phenomenon and instrument models are used to derive a proper ML framework for lifetime estimation. Unlike usual pointwise approaches, a neighborhood-wise approach is proposed to take explicitly into account the spatial correlation of data [15] . The application to real biological data allowed us to prove the spatial localisation of interactions, a new result which was not achievable with conventional methods. For future work, the main challenge is to extend the framework to deal with multi-exponential decay estimate and adaptive neighborhoods, a challenge we need to address for a large class of biological studies.

Reference: [15]

Partners: A. Chessel (University of Cambridge, UK), F. Waharte and J. Boulanger (UMR 144, PICT IBiSA, CNRS-Institut Curie)

Figure 6. Example of typical Time-Correlated Single Photon Counting (TCSPC) FLIM data. Total fluorescence intensity is shown in the center and corresponds to the sum of fluorescence intensities along the time axis at each pixel. The four side graphs correspond to time dependent photon counts in four different regions with variable sizes. By considering large regions, we observe an exponential decreasing along the time of fluorescence lifetime (see D). A: one pixel region; B and C: $3\times 3$ patches at different locations; D: $15\times 15$ patch and liftetime estimation by least-square fitting.