A deep-learning-based method to spectrally separate overlapping fluorophores based on their fluorescence lifetime

The simultaneous labelling and imaging of different bio-molecules are required to understand the relationships between the various sub-cellular components and macro-molecular complexes constituting a cell. In fluorescence microscopy, a careful selection of fluorophores is required to prevent spectral overlap, which limits the number and types of fluorophores that can be simultaneously used. This limitation can be overcome with the fluorescence lifetime, able to separate the fluorescence signal. In this study, the authors used deep learning to separate two fluorophores based on their fluorescence lifetime, taking advantage of non-linear spatial-temporal information. The training was carried out on synthetic images, and the results were evaluated on test synthetic images.