Multiway Decomposition Followed by Reconvolution of Fluorescence Time Decay Data

The use of time-resolved emission spectroscopy (TRES) is increasing as these instruments become more available, and the prices are decreasing, especially with the development of cheaper LED light sources. In this article, we propose a new methodology for analyzing TRES data. It combines two existing methods: PARAFAC and reconvolution. PARAFAC is a soft modeling curve resolution technique which has been extensively applied to steady-state fluorescence data, and reconvolution is the most common method for fitting TRES data. The proposed method is compared to two well-established methods of analyzing these data, namely, global reconvolution and tail fitting. In addition, we compare our approach with the SLICING method proposed in 2021 by Devos et al. which is also based on a soft model, but does not include the reconvolution step. All of these methods follow the assumption that the measured fluorescence signal is a linear combination of the underlying fluorophores. The comparison is based on a measured TRES data set with a mixture of three fluorophores and two sets of simulated data sets with up to four fluorophores. The results show that global fitting works well as long as the signal-to-noise ratio (SNR) is high (more than 15 dB), independent of the spacing between the emission peak maxima. SLICING does not give as good estimates of the time decay, mainly due to the challenge of defining the tail. Our proposed method gives robust and accurate results, outperforming the other techniques in cases with broad instrument response functions and high noise levels with SNRs down to 5 dB.