Wide-field time-gated SPAD imager for phasor-based FLIM applications

Abstract

We describe the performance of a new wide area time-gated single-photon avalanche diode (SPAD) array for phasor-FLIM, exploring the effect of gate length, gate number and signal intensity on the measured lifetime accuracy and precision. We conclude that the detector functions essentially as an ideal shot noise limited sensor and is capable of video rate FLIM measurement. The phasor approach used in this work appears ideally suited to handle the large amount of data generated by this type of very large sensor (512×512 pixels), even in the case of small number of gates and limited photon budget. , 1 2 2 1 12 tan 2 cal i cal theo cal cal cal cal cal

Fluorescence lifetime imaging with a megapixel SPAD camera and neural network lifetime estimation

“…Alternatively, fast visualisation methods such as phasor analysis have been proposed 46 , and have been successfully used for time-gated SPAD array FLIM data analysis 47 . In addition to the above-mentioned numerical approaches, advances in machine learning (ML) methods 48 has enabled researchers to utilise deep learning (DL) frameworks to extract the exponential decay time and component fraction information from FLIM data rapidly and without fitting 49,50 .…”

Wide-field Fluorescence Lifetime Imaging Microscopy with a High-Speed Mega-pixel SPAD Camera

“…We note that we pre-process the data by performing background subtraction and pile-up correction, where the latter is accounted for, following equation 1 in Ref. [35] and adopting the same nomenclature:…”

“…However, LSQ minimisation-based lifetime estimation is typically very demanding computationally, even with the reduction of computational times provided by Graphical Processing Units (GPUs) [33]. Alternatively, fast visualisation methods such as phasor analysis have been proposed [34], and have been successfully used for time-gated SPAD array FLIM data analysis [35]. In addition to the above-mentioned numerical approaches, advances in machine learning (ML) methods [36] has enabled researchers to utilise deep learning (DL) frameworks to extract the exponential decay time and component fraction information from FLIM data rapidly and without fitting [37,38].…”

Direct Photon-by-photon Analysis of Time-Resolved Pulsed Excitation Data using Bayesian Nonparametrics

“…1,2 Some fluorescence approaches-such as confocal microscopy, 3 two-photon microscopy 4 and super-resolution widefield applications 5 -use constant illumination to provide information on chemical kinetics, [6][7][8] diffusional dynamics [9][10][11] or spatial locations of molecules. 12,13 Other methods use pulsed illumination [14][15][16][17][18] or illumination intensity modulated at a fixed frequency [18][19][20][21][22][23] where photon arrival times encode critical information, say, on the excited state lifetime or the number of different chemical species. This is the basis of lifetime imaging 13,[24][25][26][27] that has been used to reveal information on local pH, 28,29 oxygenation 28 and other cellular metabolic traits 23,30 affected by cellular microenvironments.…”

“…This ability to treat models themselves as random variables is the key technical innovationthat prompted the development of BNPs in the first place. BNPs make it possible to avoid the computationally infeasible task of first enumerating and second comparing all models for any associated parameter values to all other competing models and their associated parameter values.The BNP approach to tackling lifetime image analysis that we propose here cannot replace phasor analysis20,23,53,57,59,62,93 or TCSPC2,14,29,32,46,81 for simple one component systems on account of their computational efficiency. However, at an acceptable computational cost, BNP approaches provide a powerful alternative.…”