Fig. 1. Number of papers focusing on FLIMfor in vivo imaging

Fig. 2. Principle diagram of TD-FLIM and FD-FLIM

Fig. 3. Imaging speed improving of TCSPC-FLIM by optimizing detection methods. (a) Using a collection module with a shorter dead time to record multiple photons within a single pulse cycle^[36]; (b) using a photon'Spinner' to shorten the dead time of the capture card^[37]; (c) using parallel array detection modules to improve photon counting rate^[38]

Fig. 4. Imaging speed improving of TCSPC-FLIM by optimizing scanning methods. (a) Implementing fast scanning with arbitrary shapes based on 2D-AOD^[39]; (b) using a multifocal array to achieve fast scanning^[41]; (c) using line excitation to realize fast scanning, and the imaging results of two slices shown on right^[44]

Fig. 5. Schematic of TG-FLIM system based on light-sheet^[55]. (a) Schematic of the system; (b) imaging results of a living zebrafish

Fig. 6. SC-FLIM technique for rapid imaging. (a) Single particle tracking FLIM based on double helix point spread function engineering^[67]; (b) implementation of high-speed FLIM using compressed sensing-based DMD spatial encoding technology^[68]

Fig. 7. Widefield FD-FLIM based on modulation of LED^[70]. (a) Schematic of the system; (b) fluorescence intensity images of Fluorescein-Glycerol (Fl-Gly) solutions with different ratios; (c) corresponding fluorescence lifetime images

Fig. 8. Correcting motion artifacts for in vivo FLIM through post-processing. (a) Inter frame artifact correction based on normalized cross-correlation algorithm^[77]; (b) inter frame and intra frame artifact correction based on the Lucas-Kanade framework^[78]

Fig. 9. Flowchart of ADCG algorithm and the fitting results with photon number of 45^[86]

Fig. 10. Three types of DL algorithms contribute to reduce FLIM processing time. (a) Traditional curve fitting is substituted by either a 3D-CNN^[92] (left) or a 1D-CNN^[93] (right) framework; (b) LLE and NIII sub networks are utilized to produce data with high photon counts from low photon counts^[96]; (c) DL algorithms are employed to improve imaging resolution^[97]

Fig. 11. Schematic of instant FLIM technique and its in vivo imaging results of neurons in mouse brain^[71]

Fig. 12. TCSPC-FLIM for living mice. (a) Imaging brain vessels and Aβ plaque in AD mice^[103]; (b) quantitative characterization of intracellular glucose concentration in cortical neurons^[104]; (c) three photon FLIM of cerebral vessels^[105]; (d) imaging of different organs in mice based on microendoscopy^[106]

Fig. 13. Applications of FLIM for intraoperative tumor diagnosis. (a) Recognition of tumor boundary based on PS-FLIM^[108]; (b) high-resolution recognition of tumor cells based on TCSPC-FLIM^[115]; (c) fast FD-FLIM with a large field-of-view based on 5-ALA labeling^[117]

Table 1. Performance comparison of several fast FLIM technologies based on DL