2.1.

KTN-EOD Setup

A DC power supply (GPR-30H100, GW Instek) and linear amplifier (Trek 2100HF, Advanced Energy) were connected in series with the electrodes of the KTN-EOD (KSCHRM0850-00, NTT Advanced Technology Corporation) to provide a bias voltage and high-frequency drive voltage, respectively. A function generator (DDS Signal Generator, Koolertron) was connected to the input of the amplifier to provide the sinusoidal scanning waveform. The temperature of the KTN-EOD entrance facet was monitored with a thermal camera (A655SC, Teledyne FLIR). The thermoelectric cooler on the KTN-EOD was driven by a Peltier controller (TDC-1010A, Cell System). An integrated 405 nm LED illuminated the KTN-EOD crystal during scanning.

2.2.

Stroboscopic Measurements of Resolvable Spots

This section describes the setup for the investigation of maximizing the speed of deflection (Sec. 3.2). A polarized beam from a 60 ps pulsed triggerable 970 nm semiconductor laser (EPL-980, Edinburgh Instruments) was spatially filtered by a $75\mu \mathrm{m}$ pinhole and then resized to form a round, collimated beam with a $1/e^2$ diameter of 0.44 mm. After rotating the polarization to align with the KTN-EOD scan axis, the KTN-EOD was placed at the beam waist. A function generator providing a sinusoidal drive waveform was also used to generate trigger signals for the laser source. The scanning beam from the KTN-EOD was collimated with a plano-concave cylindrical lens ($f=-6.4\mathrm{mm}$), and the scan range was collimated with a compound scan lens ($f=71.4\mathrm{mm}$) consisting of two spherical doublets ($f=100\mathrm{mm}$, $f=250\mathrm{mm}$) forming an intermediate image plane. The image plane of the scanning beam was then down-magnified 3× by an image relay consisting of two spherical achromatic doublets ($f=150\mathrm{mm}$, $f=50\mathrm{mm}$) to fit on a sCMOS camera sensor (CS165MU1, Thorlabs Inc.). By changing the phase between the trigger waveform and the sinusoidal scan, different scanning angles were interrogated. The total scan range was measured as the distance on the camera between the center of mass of focal spots located at the positive and negative extrema of the drive voltage range. For each drive voltage, two to five beam profiles were recorded, distributed across the scan range depending on the extent of the achieved range. The number of resolvable spots was calculated as the ratio of the total scan range to the average FWHM of a Gaussian fit to the beam profiles in the scan dimension.

The number of resolvable spots was measured at 200, 320, 440, and 560 kHz line rate using the deflector operating conditions recommended by the factory (temperature controller setpoint of 31.2 C, DC bias voltage of 218 V, and sinusoidal drive voltage amplitude of 270 $V_{\mathrm{pp}}$). In addition, the number of resolvable spots was measured at line rates from 200 to 1040 kHz in 120 kHz increments under optimized deflector operating conditions for each scanning frequency. To find the optimal operating conditions for each line rate, the DC bias voltage, Peltier set temperature, and drive voltage amplitude were iteratively adjusted to maximize the number of resolvable spots. In each case, the camera was translated along the optical axis to ensure that it was precisely positioned with the beam waist in the scanning dimension.

2.3.

Influence of Laser Average Power on Performance

This section describes the setup for the investigation of the effects of scanning performance as a function of average laser power (Sec. 3.1). The beam from an 80 MHz pulsed laser source (InSight X3, Spectra-Physics) was circularized and resized to produce a round mode with a $1/e^2$ diameter of 0.47 mm using a series of spherical and cylindrical telescopes as shown in Fig. 2(a). Based on our previous experience balancing excitation efficiency with scatter length during in vivo imaging of GFP variants,^28,29 we chose to study deflector performance at a central wavelength of 960 nm. The laser pulses were compressed using the DeepSee internal compression offered by the Insight X3 in conjunction with a home-built single prism (H-ZF72A, United Crystals) compressor.³⁰ Compression was verified by tight-focusing the beam onto a fluorescent slide in the KTN plane and maximizing fluorescence intensity by varying prism separation using the single prism compressor. After verification of compression, the KTN-EOD was placed at the beam waist. An image of the scan range was relayed to the surface of a sCMOS sensor to monitor the beam profile (see Sec. 2.2 for details).

Fig. 2

Stabilized deflection range across typical imaging powers for in vivo 2P imaging. (a) Experimental setup for capturing deflector scan profiles and thermal images, showing beam conditioning optics, KTN-EOD, and thermal camera. Inset shows optics used to relay the scanning beam to the camera. (b) Camera images of the scanning beam at each laser power using factory deflector settings. (c) Camera images of the scanning beam at each laser power using deflector settings manually optimized for each laser power. (d) Mean temperature deviation of the crystal at each laser power as measured from thermal imaging of the crystal face. (e) Relative scan range at each laser power compared to the 50 mW, optimized parameters case. EOM, EO modulator; $\lambda /2$, half-wave plate; BD, beam dump; PBS, polarizing beam splitter; CAM, camera; ZnSe, zinc selenide.

For each scan profile measurement, the KTN-EOD was temperature-controlled and was provided with a DC bias and AC drive voltage. The beam was initially blocked, and a thermal camera image of the KTN crystal face was captured, serving as a baseline temperature measurement. The incident laser power was adjusted using a combination of a half-wave plate and PBS power control module, as well as a Pockels cell (350-80, Conoptics Inc.). The power was measured at the KTN plane before each measurement with an optical power meter (S415C, Thorlabs Inc.). The beam was then unblocked and allowed to transit the crystal while the KTN-EOD was scanning. After 1 min, an image of the scanning beam and a thermal camera image were captured. A set of measurements were performed using factory recommended deflector operating conditions (see Sec. 2.2) at laser powers of 50 to 350 mW in increments of 50 mW. These measurements were repeated for the same laser powers, with a corresponding reduction in the temperature controller set point for each measurement. The new temperature controller set point for each laser power was selected to achieve a scan range equivalent to the scan range observed at low laser power (50 mW).

2.4.

Influence of Laser Peak Intensity on Performance

This section describes the setup for the investigation of the effects of scanning performance as a function of laser peak intensity (Sec. 3.1). Images of the scanning beam were captured at a low power (10 mW) and a high power (350 mW) in the same experimental setup as in Sec. 2.3. In this instance, the $1/e^2$ beam diameter at the entrance aperture of the KTN-EOD was measured to be $\sim 0.45\mathrm{mm}$. At both low power and high power, scan range data were measured using compressed and chirped pulses. The compressed laser pulse duration was determined by minimizing the pulse duration after the KTN via manipulation of our single prism compressor (see Sec. 2.3) and was measured directly via second harmonic generation frequency-resolved optical gating (SHG-FROG^30–33) to be 118 fs. Chirped pulses were created through the inclusion of 3 sets of two 5 mm uncoated zinc selenide (ZnSe) windows (WG71050, Thorlabs Inc.) in Brewster pairs (to compensate lateral Snell shifts and minimize reflective losses). Using the measured pulse spectrum and phase as the input, the theoretical group delay dispersion (GDD) and third-order dispersion from the ZnSe windows were added to simulate the resulting chirped pulse duration using PyNLO^34,35 to be $\sim 565\mathrm{fs}$. Due to reflective losses from the addition of the ZnSe windows, the highest available power for the chirped pulses was reduced to 335 mW.

2.5.

2D KTN-EOD Relay Module

To incorporate the KTN-EOD into our custom relay module, the small plano-concave cylindrical lens affixed near the exit aperture by the manufacturer was removed. The relay module shown in Fig. 5 collimates the scanning beam of KTN-EOD 1 in the $x$-dimension and relays a conjugate plane downstream using a compound scan lens ($x$-conjugator) consisting of three NIR-coated $f=50\mathrm{mm}$ spherical achromatic doublets (45-803, Edmund Optics). The scanning beam (which begins focusing in the $y$-dimension after the scan lens of KTN-EOD 1) is then collimated with an $f=19\mathrm{mm}$ cylindrical plano-convex singlet ($y$-collimator) (CKX019AR.16, Newport Corporation). After a 90 deg polarization rotation via half-wave plate (WPHSM05-980, Thorlabs Inc.), KTN-EOD 2 is placed at the conjugate plane of KTN-EOD 1 with electrodes oriented orthogonally to KTN-EOD 1. The same lenses rotated by 90 deg were positioned after KTN-EOD 2 in the same arrangement to create a second conjugate plane common to each KTN-EOD at the exit of the scanner relay module. Fine positioning of the KTN-EODs was enabled by a triple-axis micrometer translation stage (PT3, Thorlabs Inc.).

2.6.

Measurement of 2P Point-Spread Function

Laser light (960 nm, InsightX3) was delivered to a single-prism compressor (PBH71) to compensate for excess GDD beyond the compensation offered by the internal compressor of the ultrafast laser. The KTN-EOD scanner relay module and a redundant set of galvanometer mirrors (6215H, Cambridge Technology) were incorporated into a homebuilt 2P microscope (MIMMS 2.0 design; Janelia Research Campus, HHMI). An up-collimating telescope was used prior to the compressor to enlarge the Rayleigh range and ensure a collimated beam throughout the pulse compressor. The beam was then down-collimated and reshaped through a series of three telescopes before entering the KTN-EOD relay module (see Sec. 2.5) with a round collimated beam size of 0.5 mm $1/e^2$ diameter. Following the KTN-EOD relay module, the scanning beam was enlarged and scan range conjugated to a plane between the galvanometer mirrors through a 5× magnification relay. The two compound relay lenses consisted of two $f=150\mathrm{mm}$ spherical achromatic doublets (49-362, Edmund Optics) and two $f=750\mathrm{mm}$ spherical achromatic doublets (67-336, Edmund Optics), respectively. A compound scan lens consisting of three spherical achromatic doublets $f=150\mathrm{mm}$ (AC508-150-B Thorlabs) and Nikon tube lens ($f=200\mathrm{mm}$) form a final 4× afocal relay to the back-pupil of the 25× objective (XLPLN25XWMP2, Olympus), ultimately resulting in an effective NA of $\sim 0.8$. Epi-detected fluorescence signal photons were reflected into a detection arm by a long-pass primary dichroic (FF705-Di01, Semrock) and further filtered by a band-pass emission filter (FF01-520/70, Semrock) before being detected by a photomultiplier tube (PMT2101, Thorlabs).

Fluorescent beads with a $0.2\mu \mathrm{m}$ diameter (F8811, ThermoFisher Scientific) were used to measure the point-spread function and FOV of the 2D KTN-EOD microscope (Fig. 6). A $z$-stack of images with a $13\mu \mathrm{m}$ by $13\mu \mathrm{m}$ FOV (177 pixels by 40 lines) was acquired with KTN-EOD line rate of 560 kHz using ScanImage (MBF Bioscience). A motorized $z$-stage (PLS-X, Thorlabs Inc.) was used to manually translate the focal plane in $2\mu \mathrm{m}$ increments between each image acquisition. Image stacks were processed using custom Python scripts. Briefly, bidirectional scanning offsets were corrected in post processing to ensure correct registration of the 2D scanning. Then, 10,000 frames were averaged to generate a single image for each $z$-position. The average projection images for all $z$-positions were then scaled (using bicubic interpolation) to generate isometric 3D voxels ($0.068\mu \mathrm{m}$). Fluorescence profiles were extracted using this 3D volume along the respective axes with the FWHM being calculated directly from the profiles.

2.7.

In Vivo Voltage Imaging

All animal procedures were carried out in accordance with protocols approved by the University of Minnesota Animal Care and Use Committee. A circular (2.5 mm diameter) craniotomy was made above the primary visual cortex (V1; 0.5 mm anterior, 2.7 mm lateral of lambda, $400\mu \mathrm{m}$ deep) of a C57BL6 mouse. Cortex was injected with 150 nL AAV1-EF1a-DIO-ASAP3 ($5\times {10}^{12}\text{titer}$; 132318, Addgene) and AAV9-CamKII-Cre ($2.1\times {10}^9\text{titer}$; 105558, Addgene) to sparsely label layer 2/3 neurons with ASAP3. A window (triple #1 coverglass 2.5/2.5/3.5 mm diameter) was fixed to the skull using dental adhesive. A head fixation metal bar (with a metal loop surrounding the window) was also implanted posterior to the window using dental acrylic. Imaging of spontaneous activity was performed on an awake and head-fixed mouse. A reference image was taken of a neuron using the redundant galvanometer mirrors ($158\mu \mathrm{m}$ by $158\mu \mathrm{m}$ FOV) at a frame rate of 1.7 Hz. High-speed imaging was then performed using the KTN-EODs (Fig. 6). For the in vivo experiment, the 5× magnifying relay in Fig. 6 (RL3 and RL4) was replaced with a 2.44× magnifying relay to produce an FOV of $27\mu \mathrm{m}$ by $10\mu \mathrm{m}$ (64 pixels by 32 lines), at the cost of reducing the effective to $\mathrm{NA}\sim 0.39$ (estimated lateral FWHM: $0.93\mu \mathrm{m}$; estimated axial FWHM: $10.3\mu \mathrm{m}$). The first compound relay lens (RL3) consisted of $f=750\mathrm{mm}$ (Edmund Optics, 67-335) and $f=100\mathrm{mm}$ (49-360, Edmund Optics) spherical achromatic doublets. The second compound relay lens (RL4) consisted of $f=750\mathrm{mm}$ (67-335, Edmund Optics) and $f=300\mathrm{mm}$ (AC254-300-B, Thorlabs Inc.) spherical achromatic doublets. The cell membrane of the soma was scanned with 960 nm light with intermittent short continuous imaging bouts at either 14.7 kHz for 1 s or 15.6 kHz for 0.9 s repeated every 5 s. Power at the sample during imaging bouts was 39 mW. The minor differences in frame rate from session to session reflected instability in the number of lines the microscope control software (ScanImage, MBF Bioscience) allotted for flyback. Images were processed using custom MATLAB and Python scripts to extract time courses. Images were motion corrected using cross-correlation³⁶ and regions were drawn manually over the cell membrane to extract ASAP3 fluorescence traces from three separate neuron somas recorded in one mouse. $\mathrm{\Delta }F/F_0$ traces were calculated using an average of the first 50 ms of each imaging trial as the $F_0$ fluorescence value. Estimates of frame-to-frame $\mathrm{\Delta }F/F_0$ noise ($\sigma$) for each neuron were calculated from the average of the spectral density between 0.25 and 0.5 times the Nyquist frequency.³⁷ Equivalent $\mathrm{\Delta }F/F_0$ noise at 1 kHz was calculated by assuming averaging of Poisson noise such that ${\sigma }_{1\mathrm{kHz}}={\sigma }_{\mathrm{fr}}/\surd (\mathrm{fr})$, where fr is the imaging frame rate in kHz. To calculate bleaching time constants, trial traces were first normalized by dividing by the average of the first 20 frames of each trial. Decay time constants ($\tau$) were then estimated by least-squares fitting of a single exponential decay model to the trial-averaged normalized trace. Normalized signal after 1 s of imaging was estimated from the decay time constant of each neuron.

3.1.

Deflection of Femtosecond Infrared Pulses with Average and Peak Energies Typical of In Vivo 2P Imaging

In the context of in vivo 2P laser scanning, the KTN-EOD is subject to higher laser average powers than the specimen due to losses from downstream optics. Therefore, we investigated the performance of the KTN-EOD at a range of laser average powers extending above in vivo limits (up to 350 mW) [Fig. 2(a)]. As laser power was increased, the scan range of the KTN-EOD decreased and the beam profile became distorted as shown in Figs. 2(b) and 2(e). However, by reducing the set temperature on the Peltier temperature controller (see Table S1 in the Supplementary Material), we were able to recover the lost range [Figs. 2(c) and 2(e)]. Thermal camera measurements indicated that at factory-defined controller settings, increases in laser power result in uncompensated heating of the KTN crystal [Fig. 2(d)]. Optimization of the temperature controller settings reduced the observed heating, especially at the highest tested laser powers [Fig. 2(d)].

It has been reported that ultrashort pulses can cause beam profile deformation in KTN-based scanners.²³ To verify that this deflector was suitable for in vivo 2P imaging, we performed deflection tests using femtosecond infrared laser pulses (see Secs. 2.3 and 2.4) across a range of peak intensities [Fig. 3(b)]. We measured the deflector scan range at high and low laser powers for both compressed pulses and positively chirped pulses, respectively (see Sec. 2.4). Our laser pulses were compressible to $\sim 118\mathrm{fs}$ FWHM using our external prism compressor. Positively chirped pulses were achieved by the insertion of bulk ZnSe, which broadened the pulses to $\sim 565\mathrm{fs}$ (see Sec. 2.4). These pulse durations (along with the measured power and known rep-rate) were then used to approximate the peak intensity of the laser pulses [Fig. 3(b)]. We observed that although increasing average laser power was seen to reduce the scan range (which could then be recovered by temperature controller optimization), peak intensity did not have a direct effect on deflector performance (Fig. 3). When holding average power nearly constant and changing the peak intensity by a factor of 4.8 we saw no significant change in the performance. These results suggest that when filling the aperture of the KTN-EOD, ultrafast nonlinearities of KTN are negligible at pulse parameters typical for in vivo 2P imaging.

Fig. 3

Intensity dependent deflector performance at a range of average powers at or above those typical for in vivo 2P imaging. (a) Beam profiles observed at different laser powers and pulse durations. (b) Scan range percentage for each combination of pulse duration and laser power (case 1: 10 mW, 565 fs; case 2: 10 mW, 118 fs; case 3: 335 mW, 565 fs; case 4: 350 mW, 118 fs) relative to 10 mW, 118 fs (case 2). Horizontal axis shows the calculated optical peak intensity for each case.

3.2.

Maximizing the Speed of Deflection

We next characterized the deflector performance across a range of line rates using a triggerable pulsed laser and camera [Fig. 4(a)]. The triggerable source enabled us to make stroboscopic measurements of scan range and estimate the number of resolvable spots. The KTN-EOD was designed to operate at a line rate of 200 kHz, but we found that it could be operated at higher rates. Using factory recommended operating parameters, performance degraded as the line rate was increased above specification [Figs. 4(b)–4(d)]. However, it was possible to recover performance for a range of higher line rates by tuning the operating parameters (see Sec. 2.2 and Table S2 in the Supplementary Material). The deflector provided up to 32 resolvable spots at a line rate of 560 kHz with optimized operating parameters. Thermal measurements revealed that when driving beyond 560 kHz, a pronounced temperature gradient across the crystal appears, after which the performance cannot be recovered by control parameter optimization (Fig. S1 in the Supplementary Material). Due to the strong temperature dependence of the relative permittivity,³⁸ a more sophisticated thermal management strategy will likely be needed to operate reliably at higher frequencies.

Fig. 4

Deflector performance at line rates exceeding manufacturer specification. (a) Experimental setup for measuring the number of resolvable spots at different line rates. Inset shows optics used to relay the scanning beam to the camera. (b) Camera images of the scanning beam demonstrating how optimization of deflector parameters can recover performance at line rates exceeding the manufacturer specifications. (c) Relative scan range achieved for each line rate with factory deflector operating parameters and manually optimized parameters. (d) Number of resolvable spots provided by the deflector for each line rate with factory deflector operating parameters and manually optimized parameters.

3.3.

Design and Implementation of a 2D EO 2P Microscope

As opposed to conventional laser scanning technologies that produce a scanning collimated beam, the KTN-EOD produces a scanning beam that is strongly focused in the scanning dimension [Figs. 1(a)–1(c)]. A 2D KTN-EOD relay module compensates for the uniaxial focusing of the first scanner to generate a collimated beam of the appropriate size at the principal plane of the second scanner. As shown in Fig. 5, the scanning beam from KTN-EOD 1 was collimated in the $x$-dimension and relayed to KTN-EOD 2 with a compound scan lens ($x$-conjugator). The scanning beam that is now focused in the $y$-dimension was then collimated with a cylindrical plano-convex singlet ($y$-collimator). After a 90 deg polarization rotation via half-wave plate, KTN-EOD 2 was placed at the conjugate plane of KTN-EOD 1 with electrodes oriented orthogonally to KTN-EOD 1. An identical set of lenses rotated by 90 deg were positioned after KTN-EOD 2 to conjugate the scanning in the $y$-dimension and collimate the scanning beam in the $x$-dimension to a common conjugate plane at the exit of the scanner relay module.

Fig. 5

Design and implementation of a 2D KTN-EOD relay module. (a) Design overview of 2D KTN-EOD relay module with conjugated, orthogonally oriented KTN-EODs. Black arrows indicate laser scanning direction. $\lambda /2$, half-wave plate. (b) Projected views of beam propagation in the relay system. The first lens in the relay conjugates the scanning pivot points of both the KTNs. The second lens (cylindrical) collimates to make it a circular scanning beam. Dashed green lines are conjugate planes. (c) Photograph of the prototype microscope. Upper left inset: magnified image of first deflector during LED illumination. Superimposed red line illustrates the excitation path.

The 2D KTN-EOD microscope required several peripheral beam conditioning telescopes, as shown in Fig. 6(a) (see Secs. 2.3 and 2.6). A peripheral single-prism pulse compressor was required in addition to the internal compensation of the ultrafast laser (see Sec. 2.6) to accommodate the GDD of the 2D KTN-EOD scanner relay module. The GDD of the KTN crystal was approximated in its operating mode by maximizing second harmonic generation signal from a beta barium borate crystal before and after its introduction to the path via the manipulation of prism separation. Based on the difference in separation, the predicted difference in compensated GDD required to re-compress the pulses was $\sim |4500{\mathrm{fs}}^2|$, providing an estimate for the KTN.³⁹ Accounting for the GDD of the other scan optics from their Sellmeier curves, the total GDD of the 2D-KTN-EOD scanner relay was $\sim \mathrm{15,000}{\mathrm{fs}}^2$.³⁵ An up-collimating telescope was used to enlarge the Rayleigh range during transit through the single prism compressor, and a three-lens cylindrical telescope and down-collimating telescope were used to correct for astigmatism and place the beam waist downstream. Due to the relatively short Rayleigh range of a 0.5 mm $1/e^2$ diameter beam, a final down-collimating telescope was required immediately before the 2D KTN-EOD relay module to ensure that the aperture was filled with a collimated beam, where a half-wave plate ensures that the polarization axis of the laser is aligned to the scan-axis of the KTN-EOD. Following the 2D KTN-EOD relay module, the 2D scanning beam was magnified by 5×, with its conjugate plane relayed onto a set of redundant galvanometer mirrors. This redundant set of scanners allowed for imaging of larger FOVs to identify targets of interest to interrogate with the smaller FOV afforded by the KTN-EOD scanning. The conjugate plane at the galvanometer mirror pair was then relayed one final time to the back aperture of our objective, underfilling to an approximate NA of 0.8.

Fig. 6

Implementation and optical performance characterization of a 2D EO 2P microscope. (a) Layout of complete optical path including light source, beam conditioning optics, scanners, optical relays, and detection scheme. (b) Image of a single $X-Y$ plane of $0.2\mu \mathrm{m}$ diameter fluorescent beads when imaging via the 2D KTN-EOD scanners. (c) Vertical (cyan) and horizontal (red) line-outs of the intensity profile and FWHM determination of a selected bead. (d) Image of an $X-Z$ slice of the bead, and (e) axial (green) line-out of the intensity profile and FWHM determination. EOM, EO modulator; $\lambda /2$, half-wave plate; KTN-X/KTN-Y, KTN-EOD crystals; RL1, $x$-conjugator; CL1, $y$-collimator; RL2, $y$-conjugator; CL2, $x$-collimator; RL3/RL4, relay lenses; GG, galvanometer scanning mirror pair; SL, scan lens; TL, tube lens; PMT, photomultiplier tube.

A $z$-stack of images was acquired from a fluorescent bead sample ($0.2\mu \mathrm{m}$ diameter) to measure the PSF and FOV of the 2D-EOD microscope (see Sec. 2.6). The FOV was measured to be $13\mu \mathrm{m}\times 13\mu \mathrm{m}$, and the lateral PSF FWHM was found to be 0.59 and $0.66\mu \mathrm{m}$ in the horizontal and vertical dimensions, respectively [Figs. 6(b) and 6(c)]. Along the optical axis, the $z$-PSF was found to have an FWHM of $2.8\mu \mathrm{m}$ [Figs. 6(d) and 6(e)]. The deviation of the measured focal volume from that of a diffraction limited 2P focal volume at 0.8 NA (lateral FWHM: $0.45\mu \mathrm{m}$; axial FWHM: $2.25\mu \mathrm{m}$) is likely due in large part to imperfections in the compact optical relays built from off-the-shelf components.

3.4.

Validation: In Vivo 2P Imaging of a Voltage Indicator at >14 kHz

To test the ability of the 2D KTN-EOD microscope to conduct cellular resolution 2P imaging in vivo, we expressed a 2P-compatible genetically encoded voltage indicator (${\mathrm{ASAP}3}^6$) in layer 2/3 neurons of the primary visual cortex (V1) of adult mice. Beam magnification after the KTN-EOD relay module was reduced, which resulted in an effective NA of $\sim 0.39$ (corresponding to expected PSF of lateral FWHM: $0.93\mu \mathrm{m}$; axial FWHM: $10.3\mu \mathrm{m}$) to obtain a larger $x$-axis FOV (see Sec. 2.7). Neurons expressing ASAP3 were identified in head-fixed mice using standard 2D galvanometer imaging [Figs. 7(a)–7(c)] and then imaged at either 14.7 kHz or 15.6 kHz frame rates using the 2D KTN-EOD light path [Figs. 7(d)–7(f)]. To assess ASAP3 signal quality and bleaching rate during 2D KTN-EOD imaging, imaging was carried out in $0.9$ to 1.0 s bouts interleaved with 4.0 to 4.1 s dark periods that allowed the signal to recover from bleaching. High frequency noise during imaging bouts was $0.058\pm 0.007$ $\mathrm{\Delta }F/F_0$ and signals decayed with a bleaching time constant of $0.35\pm 0.03\mathrm{s}$ [Figs. 7(g) and 7(h)]. For comparison with another ultrafast scanning approach,¹⁰ we estimated (see Sec. 2.7) the equivalent noise after averaging down to 1 kHz sampling ($0.015\pm 0.002$ $\mathrm{\Delta }F/F_0$) and the normalized signal after 1 s of imaging ($0.74\pm 0.02$). The estimates indicate that net emission and bleaching during 2D KTN-EOD imaging of ASAP3 were not far from those obtained using other kHz scanning approaches.¹⁰ Due to the fact that the imaging FOV was small relative to brain motion, the correction of the signal for motion artifacts was particularly challenging and sparse deflections in the ASAP3 signal could not be confidently classified as either artifacts or action potentials. Nonetheless, these results demonstrate that the 2D KTN-EOD microscope was capable of cellular imaging in vivo with high spatial and temporal resolution.

Fig. 7

In vivo characterization of 2D EO microscope bleaching kinetics. (a)–(c) 2P images of three layer 2/3 pyramidal neurons expressing ASAP3 collected using standard scanning galvanometers. (d)–(f) 2D KTN-EOD scanning images (mean of 1000 frames) of the regions marked by dashed boxes in (a)–(c). Images were acquired at both [(d),(e)] 15.6 kHz and (f) 14.7 kHz frame rates. (g) Example single trials showing ASAP bleaching $\mathrm{\Delta }F/F_0$ for the three cells membrane regions indicated with the dotted lines in (d)–(f). SNRs ($\sigma$) are indicated in the corresponding legends. (h) Trial-averaged normalized fluorescence exponential decay bleaching kinetics (shading indicates mean ± standard error of the mean) were calculated for all three cells (cell 1 $N=106$ trials; cell 2 $N=118$ trials; cell 3 $N=117$ trials; $N=1$ mouse). Exponential decay fits to the trial averaged data are shown with the solid lines and time constants ($\tau$) are provided for each of the three cells in the respective legends. The objective back-aperture was underfilled to an effective NA of $\sim 0.39$ (with a theoretical PSF of lateral FWHM: $0.93\mu \mathrm{m}$; axial FWHM: $10.3\mu \mathrm{m}$).