The biochemical resolving power of fluorescence lifetime imaging: untangling the roles of the instrument response function and photon-statistics

Abstract

A deeper understanding of spatial resolution in microscopy fostered a technological revolution that is now permitting us to investigate the structure of the cell with nanometer resolution. Although fluorescence microscopy techniques enable scientists to investigate both the structure and biochemistry of the cell, the biochemical resolving power of a microscope is a physical quantity that is not well-defined or studied. To overcome this limitation, we carried out a theoretical investigation of the biochemical resolving power in fluorescence lifetime imaging microscopy, one of the most effective tools to investigate biochemistry in single living cells. With the theoretical analysis of information theory and Monte Carlo simulations, we describe how the "biochemical resolving power" in time-resolved sensing depends on instrument specifications. We unravel common misunderstandings on the role of the instrument response function and provide theoretical insights that have significant practical implications in the design and use of time-resolved instrumentation.

Simple and Robust Deep Learning Approach for Fast Fluorescence Lifetime Imaging

“…In terms of 𝜏𝜏 𝐴𝐴 , the performance of NLSF declined as photon counts decreased because the fitting process (Levenberg-Marquart deconvolution) required enough counts to guarantee accuracy. The provocative study [42] using similar synthetic parameters to ours unravelled that, for mono-exponential decays, the necessary number of photons of the deconvolution method is around 100 for the minimal resolvable lifetime of 0.3 ns. Moreover, pre-set initial values are critical for the deconvolution method, and we did not fine-tune them here.…”

In vitro and in vivo NIR Fluorescence Lifetime Imaging with a time-gated SPAD camera

“…Nonlinear Least Square Fit (NLSF, see Online Methods ) is a standard technique which only requires an instrument response function (IRF) measurement, as well as good signal to noise ratio (SNR) to obtain accurate results, but can be computationally demanding 25,29 . A user-friendly and fit-free alternative to NLSF analysis is the phasor approach 18,30–3434 , which is based on the computation of a pair of Fourier coefficients of the decay (Fourier harmonic n or phasor frequency f = n/T ), which are then represented as a point in the phasor plot .…”

“…We show here that these are not obstacles to accurate lifetime measurements. Indeed, the gate step size and the photon count, more than the gate duration or its rise- and fall-times, are the primary parameters affecting the measured lifetime precision 18,24,25 .…”

“…Next, most NIR dyes exhibit lifetimes far shorter than visible dyes (few hundreds of picoseconds – ps – compared to few nanoseconds). Lastly, MFLI-FRET involves quantifying fluorescence decays from two or more fluorophore species or states simultaneously, which requires larger signal-to-noise compared to mono-exponential cases, leading to additional challenges 25 .…”