Fig. 1. Schematic diagram of volumetric imaging by Duo-2P. (a) Configuration of Duo-2P. Two sets of resonant scanning two-photon microscopes are positioned along the same optical axis on opposite sides of the sample. Each microscope is responsible for volumetric imaging of half the thickness of sample layer by layer. (b) Alternate volumetric imaging by Duo-2P, one frame at a time. Z-stack images from the two microscopes are combined to synthesize a full 3D image of the sample. Frame T/M and B/M represent the frames of layers (LTn/LBn) acquired from top/bottom to middle of the sample by the two microscopes, respectively. (c) Excitation beam switching is accomplished using a Pockels cell and a polarizing beamsplitter (PBS). Epi- and contralateral fluorescent signals are collected simultaneously by the two microscopes, detected by respective PMT, and digitally combined to form a frame image. M represents mirrors here.

Fig. 2. Simulations of bidirectional emission collection in Duo-2P. (a) Enlarged schematic of the imaging chamber outlined by the dashed box in Fig. 1(a). The imaging chamber is filled with culture medium, and the upper objective is immersed in the medium. A cover glass (1#, 0.13–0.16 mm) separates the lower objective from the sample, with distilled water filling the gap between the objective and cover glass. D represents the imaging depth from the sample surface to the fluorescence point. Δd represents the defocus distance between the fluorescence point and the focal plane of the contralateral objective. Its value is defined as positive when the focal plane is above the fluorescence point. (b) Zemax simulation of collection efficiency in scattering tissues. The position of the objective, lens 1 (L1), lens 2 (L2), and detector has been optimized for the telecentric light path. The detector’s receiving area for the light is limited to a circular area of 5 mm in diameter. (c) Simulation illustrating the variation in collection efficiency (η) influenced by imaging depth (D) and defocus distance (Δd) marked in (a). The impact of coverslips on collection efficiency is ignored. The arrowed lines depict the collection efficiency variation trajectory when using Duo-2P to image an example 500-μm-thick sample. The solid line represents epi-collection efficiency, while the dashed line depicts the efficiency from the contralateral side. (d) Collection efficiency of epi-2P and Duo-2P at different imaging depths with steps of 10 μm for a 500-μm-thick sample with a scattering length of 100 μm. (e) Simulated SNR in relation with the contralateral collection and fluorescence intensity. Contralateral collection is measured with the ratio α, which is the contralateral fluorescence signal over the epi-fluorescence signal. F represents the relative fluorescence intensity. At the same SNR (the two contours), the contralateral collection ratio increases as the fluorescence intensity decreases.

Fig. 3. Excitation energy reduction by bidirectional excitation in Duo-2P. (a) Comparison of the excitation laser power for Duo-2P and epi-2P imaging with the same SNR at a sample thickness four times the scattering length (ls). The required laser power P at depth z within the sample is measured by the ratio of excitation power when imaging the sample surface. (b) Ratios of epi-2P’s excitation energy input to that of Duo-2P imaging. The blue line represents the ratio calculated for the same fluorescence intensity at different imaging depths. The points in brown are calculated using the fluorescence collection efficiency simulated in Fig. 2(c), given the same SNR at each imaging depth. The fitting dashed line depicts the trend, and the shadowed areas exceed the imaging depth limit of epi-2P.

Fig. 4. Duo-2P images of fluorescent microspheres in the tissue phantom, showing the improvement in excitation energy input, SBR, fluorescence collection efficiency, and SNR when compared to epi-2P. (a) Volumetric images of a tissue phantom containing fluorescent microspheres acquired by Duo-2P and (b) epi-2P. The total thickness of the tissue phantom is 700 μm. (c) Scattering length, obtained by fitting the logarithm of fluorescence intensity with depth. (d) Comparison of images acquired by Duo-2P and epi-2P at different imaging depths. Scale bar, 50 μm. (e) Magnified images of the signal and background for SBR calculation, indicated by the red and orange arrows, respectively. (f) Raw data’s SBR versus penetration depth, obtained using Duo-2P and epi-2P. Each individual data point on the figure represents the SBR calculation result of one microsphere. (g) Image with enhanced SNR, generated by fusing the images from epi- and contralateral collections. Scale bar, 50 μm. (h) Ratio of fluorescence collection efficiency in Duo-2P to that of epi-2P imaging versus penetration depth. (i) Statistical analysis of the SNR’s ratio of the fused images to that of the epi-2P images, with respect to the penetration depth. The statistical significance is determined using one-way ANOVA with the Tukey’s multiple comparisons test. ns, not significant (p>0.05). ***, p<0.001. Data were presented as mean±standard deviation here. (j) Statistical analysis of raw data’s SNR from epi-2P and Duo-2P versus excitation laser power. In (h) and (j), each individual data point was plotted on the corresponding box graph, with a center line at the median, an upper bound at the 75th percentile, a lower bound at the 25th percentile, and whiskers extending to the minimum and maximum values.

Fig. 5. Continuous volumetric calcium imaging of Nms neurons in the SCN slice. (a), (c), (e) Time averaged images of layers T12 (the 12th layer from the top), B25 (the 25th layer from the bottom), and B9 (the 9th layer from the bottom). (b), (d), (f) Excerpted baseline-corrected calcium traces of 30 representative neurons from (a), (c), (e). (g) State-switching behavior of 8 Nms neurons in the SCN slice across 9000 sampling points. (h) Volumetric images of the Nms neurons in the SCN. The entire imaging volume is 550  μm×550  μm×380  μm (corresponding to 512×512×50  voxels). (i) Wavelet transform of three representative neurons for visualizing the activity frequencies and state-switching behavior. Scale bar, 100 μm.

Fig. 6. Schematic of the complete optical path of the experiment system. The system can be switched between epi-2P mode and Duo-2P mode for comparisons. The upper and lower microscopes share the same laser source. The Pockels cell is used to gate the laser pulse used for the excitation of both sides of the sample.