Open Access
Add to BibSonomy
Abstract
The streak camera is a picosecond resolution photodetector with parallel input capability; however, the degree of multiplexing is limited by crosstalk and temporal uncertainty in the sweeping field. We introduced a fixed time delay between adjacent fibers to reduce crosstalk in the synchroscan mode. The fixed delay and a tunable electronic delay between the input pulse and the synchroscan unit allows robust separation modes between the streaks, while spatial and temporal nonlinearities can be calibrated in. The efficacy of the design is demonstrated through a 100-fold multiplexed confocal fluorescence lifetime imaging microscope in live cells.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
The streak camera is a time domain photodetector used for pico- to femtosecond temporal detection from multiple input channels [1,2]. The steak camera is particularly useful for single-shot applications with non-repetitive dynamics such as X-ray free electron laser (FEL) pulse [3], pyrometery [4], pulsed polarimetry [5], and single-shot imaging [6]. Moreover, the streak camera has been applied to repetitive measurements in biomedical applications including fluorescence lifetime imaging (FLIM) [7–9], photoacoustic tomography [10,11], and optical biopsy [12,13]. Comparing to other time-domain detection techniques, the streak camera can measure multiple photons at once with high temporal resolution.
The streak camera achieves its temporal resolution by converting the time measurement into a distance measurement [14]. Briefly, incoming photons are converted into photoelectrons when entering the photocathode. The photoelectrons are amplified and accelerated when passing through a micro-channel plate (MCP) towards a phosphor screen along its traveling direction. A second sweeping electrical field orthogonal to the traveling direction is applied to deflect these electrons. The sweeping field varies in time, such that the deflection field applied to an incoming photoelectron depends on the original photon arrival time at the photocathode. As a result, the measured deflection distance on the phosphor screen corresponds to arrival time of the photon. During a single image acquisition many photon events are registered in the direction of the sweeping voltage, creating a “streak”. As a result, the temporal resolution of a streak camera is measured in the orthogonal deflection direction and typically in the range of femto- to picoseconds.
In a single-scan mode, the sweeping voltage is linearly applied in one direction. Deflection distance is linearly proportional to the arrival time of the incoming photon (linear time axis). The single sweep mode repetition rate is typically limited to a few MHz. To achieve a higher repetition rate, the sweeping voltage can be applied in a sinusoidal fashion in both directions [15], with the sinusoidal field is synchronized with a repetitive pulsed excitation source (synchroscan mode) achieving high repetition rate up to the GHz regime. The challenge with synchroscan mode is that the deflection distance of generated photoelectrons is no longer linear, resulting in a sinusoidal time axis.
Multiple channels can be acquired simultaneously by a streak camera; however, the number of channels is limited by crosstalk between adjacent channels [2,16]. To reduce such crosstalk in single sweep mode, we previously developed a fiber array with alternating fibers in two different lengths [17,18]. As a result, light signal from adjacent channels reach the detector at different times and spatially separates the generated streaks on the readout image. The spatial separation significantly reduces the crosstalk in adjacent channels, thus allowing a higher degree of multiplexing. In this work, we extend the application of our alternating fiber design to a streak camera with a synchroscan sweeping unit.
The spatial non-uniformity of the electrical field contributes to temporal uncertainty of the measurements. With increased multiplexing, a much larger area of the streak field is used, artifacts from the deflection of the electrical field will lead to significant nonlinearities in the time domain. For high precision measurements, such errors need to be calibrated and corrected especially when multiplexing is used. Here, we present an alternating fiber delay scheme to minimizing the crosstalk between multiplexed input channels, while a separate fiber delay bundle allows correction and calibration under the synchroscan mode. To demonstrate the efficacy of the presented design, we coupled the fiber-optic streak detector to a multiplexed confocal microscope and used it to measure the fluorescence lifetime generated by 100 input channels. We identified asymmetry of the transient time with respect to the sweeping voltage direction and corrected for these artifacts to accurately estimate fluorescence lifetime. The combination of alternating fiber design and temporal corrections pushes the streak camera to its full temporal detection capacity.
2. Experimental design
2.1 Optical system setup
Figure 1(a) shows the schematic of the implemented design. A 40MHz pulsed diode laser (3 mW, 60 ps LDH-P-C-440M, PicoQuant GmbH, Berlin, Germany) provides excitation at 440 nm. The beam is shaped into a uniform square field and reflected by a dichroic filter (Di01-R442-25x36, Semrock, Rochester, NY) to a microlens array (MLA) (500 μm center-to-center pitch, square MLA, SUSS MicroOptics, Neuchâtel, Switzerland). A 10 × 10 foci array image generated by the MLA is then relayed to the side imaging port of an inverted fluorescence microscope (Axiovert 200, Carl Zeiss AG, Oberkochen, Germany). A 40 × oil emersion objective (Fluar 40 × /1.30 Oil, Zeiss) is used to create an image of the foci on the sample. The emitted fluorescence signal from each excitation spot is collected. The emission signal passes through the dichroic filter and an emission filter (FF01-542/27-25, Semrock); then on to a second lenslet array. It is then coupled to a 2D-1D fiber optic bundle (0.11 NA, fused silica, n = 1.44, 50μm diameter core, 125μm cladding), which guides the photons to the streak camera (SC-10, Optronis GmbH) through its 100 channel 1D end. The streak camera uses a 40 MHz synchroscan sweeping unit synchronized to the excitation source repetition rate. The readout imager used in the streak camera is a scientific CMOS sensor (sCMOS, PCO edge 5.5, pixel size 6.5μm). Confocal images are constructed by raster scanning the generated foci across the sample using a window-based scanner (Fig. 1(a), yellow bars). The window scanners shift the foci image at the microscope's conjugate plane. One scanner shifts the image along the x-axis and the other along the y-axis. The operating principle and characterization of the scanning window design can be found with more details in our previous publication [19].
Fig. 1 (a) Optical setup of the multifocal confocal streak system. (b) Schematic illustrating inward (left) and outward (right) scanning modes. The generated streaks in both modes are illustrated above, with a blue-red pseudoscale (left) representing intensity of generated streaks in that image. The sinusoidal sweeping voltage profile is shown below each image; black and white triangles indicate the start of the streaks from alternating fibers, blue and red lines indicate the near-linear section of the sweep profile, which is used to generate the streaks in the above image. The black and white arrows in the image above correspond to arrows indicated on the sinusoidal profile below.
2.2 Alternating fiber approach to reduce crosstalk
Even though photons exiting the 1D fiber are tightly focused into the photocathode, the generation of photoelectric activity broadens the readout signal from the sCMOS sensor. There are multiple causes for such broadening including: electron discharge from one channel to another at the MCP, limited spatial resolution of the MCP and phosphor screen, and diffusion of photons generated at the phosphor screen to a larger area. These effects introduce intra-streak crosstalk in a multiplexed setting and limit the number of input channels. Previously, we introduced a 4.8 ns (1 m) difference in the alternating fiber between adjacent channels to split the streak formation across the readout camera, in single sweep mode [17]. For the synchroscan streak camera, the alternating fiber length needs to take into account the period of the sinusoidal sweep and its nonlinearity.
In our current system setup, we use a 40MHz synchroscan unit, resulting in a 25 ns full sweep cycle (Fig. 1(b)). Adding a 12.5 nanosecond delay between consecutive fibers ensures that photons are split between two halves of the sweep cycle. Therefore, half the photons are collected as the sweeping voltage moves from left to right, whereas the other half are collected as the sweep moves from right to left. Furthermore, this fixed delay between alternative channels can be combined with an electronic delay between the excitation pulse and the synchroscan unit to achieve different streak separation modes. Figure 1(b) shows two of these modes which are referred to as ‘inward sweeping’ or ‘outward sweeping’ mode.
In the inward sweeping mode, the generated streaks start near the turnaround points of the sweeping voltage and are swept towards the center of the streak. In the outward sweeping mode, the streaks start at the center of the sweep and are swept towards the turnaround points. These modes offer flexibility in using the streak for different multiplexed settings. Inward sweeping extends each streak to use the full sweeping range and thus has a larger measurement time window, which is more suitable for longer temporal measurements. However, the streaks have higher amount of crosstalk as the time windows overlap. On the other hand, outwards sweeping is restricted to a shorter measurement window with reduced intra-streak crosstalk.
The crosstalk can be quantified by measuring the intensity spread of a single streak to adjacent streaks (Fig. 2(b)). We describe the crosstalk as the percentage of the maximum intensity as a function of distance away from the center of the streak (Fig. 2(a)). We measure this profile at the beginning of the fluorescence-generated streak where the highest crosstalk occurs (Fig. 2(a), black line). Figure 2(b) shows an example of the collected streak images for both inward and outwards modes. In our setup, we measured the distance between two neighboring streaks in the inward mode is about 9 ± 1 pixels (Fig. 2(b), red dashed lines for two streaks adjacent to the central streak). This corresponds to ~10% crosstalk. The distance between two streaks doubles for the outward mode (Fig. 2(a) and 2(b) cyan solid lines for two streaks adjacent to the central streak). Consequently, the crosstalk in the outward mode is reduced to <1%. Hence, we use outward for fluorescence lifetime imaging in section 4.0.
Fig. 2 (a) Normalized intensity map of a single streak. The black line shows the normalized intensity profile of the maximum intensity point on the streak perpendicular to the sweeping direction. The red dashed lines indicate the positions of the two adjacent streak to the central streak in inward scanning mode. The cyan solid lines indicate the positions of two streaks adjacent to the central streak in the outward scanning mode. This profile is used to determine the crosstalk as a function of distance away from the streak (in pixels). (b) Examples indicating corresponding location of the streaks of interest (yellow arrow) and the adjacent streaks (red dash lines and cyan solid lines). Pixel size: 6.5μm
3. Image processing and lifetime estimation
3.1 Time axis linearization for a single streak
Time measurement in the streak is recorded as a deflection distance relative to the excitation pulse. In the single sweep mode, the deflection distance of an incoming photon is linearly proportional to its arrival time. In synchroscan sweeping, the time axis follows a sinusoidal pattern. Time linearization for a single streak is achieved by first determining the distance, measured in pixels, between the turnaround points for the sweeping profile. This distance (x) can be then converted into time units using:
Where T is the sweep period, c is the sweep center and A is the sweep amplitude. These parameters are obtained by accurately localizing the position of the turnaround points for the sweeping profile for streaks generated on the sCMOS sensor. Experimentally, this is achieved by using a continuous light source. Non-synchronization of incoming photons generates line profiles where the intensity is inversely proportional to the deflection time at each pixel. Deceleration of the sweeping voltage at the turnaround points results in larger accumulation of photons, which identifies the turnaround position. For the unsynchronized light the measured intensity can be modeled as:
Sub-pixel localization of the turnaround points requires estimation of the spread profile of photons coming from a single fiber, whilst the streak camera is set in non-sweeping mode (focus mode). Attempting to linearize the time axis without taking the point spread function (PSF) into account would result in an overcorrection of the non-linearity. Taking the PSF into account, the actual profile is:
We also corrected for other standard non-uniformities such as differences in the collection efficiency across the sCMOS sensor.
3.2 Time axis linearization for multiple streaks
When using the entire image sensor to collect a large number of streaks, the effects of temporal and spatial distortions become non-negligible. Barrel distortion (spatial) and transient time (temporal) effects must be corrected to ensure the timing accuracy of the collected photons is consistent for all streaks. In this section, we will present an experimental approach to quantify and correct both of these effects.
The optics that relay the image from the phosphor screen to the sCMOS sensor introduce barrel distortion. In the presence of barrel distortion, streaks at the top and bottom of the image are collapsed into a smaller number of pixels. Since the number of pixels used to capture the streak affects the temporal resolution, barrel distortion must be corrected for a given optical configuration to ensure that every streak is spread across an equal number of pixels at the sCMOS sensor.
Barrel distortion is corrected by imaging a grid of points with known distances between them and taking into account the resulting distorted image, a transformation map is numerically calculated using the following equation:
Where (xd, yd) is the position of the point in the deformed image, (x,y) is the position of the same point in the corrected image, r is the radial distance from the center and k1,k2, and k3 are radial distortion parameters. In most cases parameters beyond k2 have negligible contribution and they are omitted [20]. Thus, the barrel distortion correction is determined by numerically solving k1 and k2 to estimate the inverse transform.
The transient time effect occurs because the sweeping voltage, at the center of the wavefront, moves slightly slower than on the edges. This effect results in the non-uniform distribution of the sweeping voltage from the center of the streak tube to the two edges. The grid mapping approach can also be used to correct for transient time effects [21]. However, the input of the streak camera does not permit imaging of a 2D grid. Instead, we simulate a grid by horizontally shifting the line of points generated by the 1D fiber input. This is achieved by simply changing the delay between the excitation source and the synchroscan unit, using fixed delay steps. The 1 × 100 line of streaks is generated during a single acquisition. Using a 510μM quenched Fluorescein dye with a very short lifetime (~17ps, using 12.2M NaI, pH > 9.0), we generate short spot-like streaks. This is equivalent to estimating the instrument response function (IRF) in the absence of fluorescence signal for each individual fiber [22]. A digital delay generator (Picosecond Delayer, Micro-Photon Devices) is used to incrementally shift the spots by 0.5ns after each acquisition. This process is repeated until the entire 25ns period is covered resulting in 50 delays steps. We locate the position of each spot by its center of mass. The generated 50 × 100 points are split into two sets of 25 × 100 based on their sweeping direction. These two sets were used to distinguish the left sweeping from the right sweeping time maps (Fig. 3(a)). All points are equidistant both horizontally (temporally governed by fixed delay step) and vertically (spatially governed by the 1D input arrangement of the fibers) and can be used to estimate the spatial and temporal distortion effects. Since the intra-distance between two adjacent fibers at the 1D input of the fiber optic is fixed, the distance between two vertically adjacent points must remain fixed for every vertical line on the simulated grid. This distance can be used in solving Eq. (4) to correct for the barrel distortion. The new grid, corrected for barrel distortion, represents the measured temporal grid for the entire streak. The full temporal map can be generated via interpolation.
Fig. 3 (a) Left and right sweep time maps generated by interpolating the simulated grid using a delay generator. (b) Lag time map generated by taking each map and subtracting the time axis for the central horizontal line. Transient time asymmetry is observed when comparing the left map to the right map. Pixel size: 6.5μm
Recall that the sweeping voltage is moving left and right in a sinusoidal motion, which means half the streaks generated appear flipped in the opposite direction (Fig. 1(b)). Temporal measurements, discussed in full detail later, indicate a discrepancy between the streaks decaying to the right compared to the left (data not shown). We generated time maps to examine this closer and correct for this effect (Fig. 3(a)). The transient time effect is caused by the lag of the sweeping wavefront at the center of the streak in comparison to the edges. We calculate a lag map by comparing every horizontal line to the central horizontal line in the generated time maps. Figure 3(b) demonstrates the asymmetric profile of the transient time effect with respect to the sweeping direction. These generated time maps can be used to extract the time axis for the streaks generated from fluorescent samples, as shown in the next section. It should be noted that these transient maps need to be measured for individual streak tubes at specific sweeping amplitude.
4. Multiplexing FLIM with streak camera
The alternating fiber delay scheme is used to collect fluorescence lifetime information from a 10 × 10 multiplexed confocal microscope. FLIM experiments were performed using two standard fluorescent dyes, a fixed Convallaria sample, as well as a live cell sample expressing a fluorescently-labeled protein using the outward sweeping mode.
4.1 FLIM imaging and fluorescence lifetime estimation
As a reference, during each experiment we collected and average 300 streak frames for a solution of 10uM Coumarin-6 dissolved in ethanol with known lifetime of 2.4 ± 0.1 ns [23], 10μM Fluorescein dissolved in 0.1M NaOH solution (pH 10) with a known lifetime of 4.1 ± 0.1 ns [24] and a 510μM quenched Fluorescein with a known lifetime of 27ps [22]. Quenched fluorescein was used to determine the IRF for each individual streak as well as their position on the sCMOS camera. To perform a FLIM scan, the 10 × 10 excitation points were raster scanned across the sample. At each step, a streak image was collected. The sCMOS integration time for each time step was 0.1s. For standard fluorescent dyes a 10 × 10 scan was acquired (10s total scan time). Each individual streak represents the fluorescence lifetime profile captured by individual foci at that point (Fig. 1(b)). Prior to extracting the lifetime decay from each streak image, background subtraction and barrel distortion corrections were performed. A rectangular region of interest (ROI) was drawn highlighting the boundary of each individual streak (Fig. 4(a)). For each ROI, the intensity profile was summed and collapsed to a single line producing the non-linearized lifetime profile.
Fig. 4 (a) Sample streaks for quenched fluorescein (IRF), Coumarin-6 and Fluorescein are shown above. Corresponding linearized lifetime decays for the average of several of these streaks is plotted below. The IRF has a second hump, which is a characteristic of the pulsed diode laser operated at high power [26]. These humps were excluded from the delay experiments used to generate the time maps. (b) Fluorescence lifetime image of Coumarin-6 and Fluorescein solutions. Images were acquired at room temperature using a 10 × 10 confocal scan. The dark squares are a result of dead fibers and were excluded from the analysis.
Fluorescent profiles were separated into left decaying and right decaying profiles. This annotation is important to determine the corresponding time map. The generated temporal maps (Fig. 3(a)) were used to extract the time axis for each individual streak. The ROI used for each individual streak was applied to the corresponding temporal map to determine the time axis. The time axis was used to linearize the extracted lifetime decays by Eq. (1). Each decay was fit to a single exponential profile by the Maximum Likelihood Estimation (MLE) algorithm [25] to determine the lifetime per pixel. We restricted the window of fitting to < 4.5 ns (Fig. 4(b)) because the decays reach a turnaround point beyond 4.5 ns, where the time axis collapses to a single point and information about the photon arrival is lost.
The streak image analysis and lifetime extraction were achieved using a custom-built C + + software. ImageJ was used to false-color the lifetime maps and extract the average and standard deviation from each map, as shown in Fig. 4(b).
4.2 Fluorescence lifetime of standard fluorescent dyes
Figure 4(a) shows examples of normalized decays for 510μM quenched Fluorescein (measured IRF), Coumarin-6 and Fluorescein solutions. Figure 4(b) shows the reconstructed FLIM images for both Coumarin-6 and Fluorescein, where lifetime is represented on a pseudo-color scale in the image and corresponding lifetime histogram (below). As expected, the estimated lifetime for Coumarin-6 was 2.5 ± 0.1ns. The estimated lifetime for Fluorescein was 4.0 ± 0.5 ns.
4.3 Confocal FLIM imaging of convallaria and live breast cancer cells
To demonstrate the confocality of the developed microscope with the streak camera as a multiplexed detector, we performed confocal FLIM imaging for a fixed Convallaria sample (Convallaria Majalis fixed sample, Leica, Wetzlar, Germany) and live MCF-7 breast cancer cells expressing a protein, Bcl-XL, tagged with a fluorophore, mCerulean3. We will refer to this protein fusion as “mCerulean3-Bcl-XL”.
MCF-7 cells expressing mCerulean3-Bcl-XL were seeded in a 384-well plate (Cell Carrier Ultra, Perkin Elmer) with 5000 cells/well and imaged 24 hours later. Bcl-XL is primarily localized at mitochondria, which are the sharp elongated features surrounding the dark (no protein expression) circular nucleus [27]. The cells were imaged on our streak system as well as on an ISS-Alba fluorescence microscope.
Imaging on the ISS was performed using a 60x water objective with a 445nm pulsed laser (20MHz, LASOS Lasertechnik GmbH) excitation and a 479-40 bandpass filter. The excitation power was measured to be 4.5µW on average at the objective. The images shown in Fig. 5 (middle and right) were acquired for a 100 × 100μm field of view using a 265 × 256 pixels (391 nm pixel size), in 45s. The confocal pinhole unit was set to be 2.5 airy disk units (middle image). On the same microscope, the confocal pinhole was removed from the optical path to obtain a widefield image.
Fig. 5 Confocal images for live MCF-7 cells expressing mCerulean3-Bcl-XL. Images of the same sample were acquired using our multiplexed streak system (left), compared images acquired on an ISS-Alba in single point confocal mode (middle) and wide field mode (right). Dark squares in the streak image indicate the positions of the dead fibers, mentioned in Fig. 4. Scalar bar: 15μm.
Confocal FLIM imaging was performed using a 40× oil objective for a 125 × 125 μm field of view divided into a 300 × 300 pixels (416 nm pixel size). The excitation power was measured to be 1.15 mW on average (~11.5 µW for individual foci). A 483-35 bandpass emission filter was used. The imaging was acquired using a 30 × 30 scan at 10 ms integration time resulting in a 90 s acquisition time. At each scan step, 100 streaks were collected. These streaks were converted into fluorescence lifetime decays following the procedure described in section 4.1. At the end of each scan, a FLIM cube (x-position × y-position × lifetime decay) was reconstructed by combining the fluorescence decays from all pixels.
As expected, compared to the standard fluorescent dyes, Convallaria and live cell samples emitted 10 times fewer photons. Estimation of lifetime from samples with poor photon economy was inaccurate on a single pixel basis. To compensate for this effect nearest neighbor binning was applied. A 2D pixel mask was generated by applying mean filtering to the intensity image prior to binning in order to exclude background pixels.
mCerulean3 is an enhanced fluorescent protein with a single-exponential fluorescence lifetime (ex/em: 433/475 nm). In solution, the protein has a lifetime of 4.1 ± 0.1ns [28], while the expected lifetime of mCerulean3 fused to Bcl-XL is 3.8 ± 0.1ns [29]. The measured lifetime in live cells expressing mCerulean3-Bcl-XL was 3.6 ± 0.3ns (Fig. 6(b)); showing a single-exponential fluorescent lifetime. Local changes in lifetime can be seen in a Convallaria sample shown in Fig. 6(a). Photon weighted lifetime map of Convallaria shows ring like structures have different lifetimes; ranging from 1.7 ns to 3.1 ns. These measurements are in agreement with previously reported values [30].
Fig. 6 Intensity weighted lifetime maps generated by fitting extracted lifetimes from data acquired on the streak microscope. Scalar bar: 15μm. (a) Confocal FLIM images of Convallaria and (b) live MCF-7 cells expressing mCerulean3-Bcl-XL. 3 × 3 binning was used for the Convallaria FLIM cube, while 7 × 7 binning was used for the MCF-7 FLIM cube.
5. Discussion and conclusion
Streak camera is one of the fastest photodetectors for measurement of ultrafast events, including up to attoseconds regime [31,32]. Besides its superior temporal resolution, streak camera can also be multiplexed to increase its throughput. Previous work by Qu et al. established the feasibility of synchroscan streak camera multiplexing [2]. In their implementation, the 1D input slit was replaced by an array of pinholes, where each pinhole generates a streak. The cross-talk depends on the intra-pinhole distance and the length of the generated streaks. Qu et al’s approach is suitable for measuring short lifetimes, because in their system the streak can only extend a short distance before contaminating the adjacent streak. Our fiber array design with alternating fiber lengths separates light signal from adjacent channels into two different optical paths leading to spatial separation on the readout image sensor. We have demonstrated how this approach extends parallel acquisition capabilities of the streak camera, equipped with a synchroscan mode, to reduce cross-talk.
The design presented here offers flexibility for imaging streaks generated for both long and short time events. The alternating difference in fiber length is chosen to be half of the period of the laser repetition rate. When combined with the synchronization delay, different sweeping modes can be selected based on the signal and the degree of cross-talk between adjacent streaks, offering additional flexibility for users. Our multiplexing scheme can be further modified to create several different operation modes. For instance, introducing more than one delay length in the array can create more intricate streak separation arrangements.
Replacing the linear sweep unit with a synchroscan sweep unit significantly improves the overall acquisition speed; however, spatial and temporal distortion must be corrected. In a multiplexed setting, the generated streaks extend across the entire image sensor. Thus, spatial non-uniformity across the sweeping field produces temporal uncertainty of the measurements. We presented two crucial steps to ensure that temporal measurement conditions are consistent for all the streaks. First, non-synchronous light is used to locate the exact position of the turnaround points for the sinusoidal sweep profile. Second, temporal asymmetry in the sweeping field caused by lag of the field front at the center of the sweep in comparison to the edges is identified. This asymmetry depends on the sweeping direction. We simulated a grid by temporally shifting a 1D array of points while imaging a quenched fluorescein sample, to compute temporal maps used to extract the time axis. Alternatively, if quenched fluorescein is not within the desired excitation/emission window, direct reflection of the excitation light could be used to generate the same temporal map. Here we have also used this simulated grid to extract radial coefficients and correct for barrel distortion effects. Altogether these corrections are essential in a highly robust multiplexed synchroscan streak camera.
To demonstrate the efficacy of the design, we coupled the fiber-optic streak detector to a multifocal confocal microscope and used it to measure fluorescence lifetime images. FLIM images acquired with the streak system is comparable to those acquired with a commercial frequency domain confocal FLIM instrument in confocal and wide field mode. Looking at the sharpness of structures in the cell, the images taken on the streak system are superior to those taken in wide field mode, however, the ISS-Alba confocal images have better image quality. We have measured the full width half maximum profile of a fluorescent bead to be 510 nm (data not shown). We attribute this broadening in the point spread function to square lenslet aperture effects [33] as the MLAs used in this setup do not have a circular aperture masking applied to their surface.
The current MLA coupling efficiency to the fiber optic bundle is not optimized to fully reject out of focus light. Fluorescence spots imaged onto the fibre bundle are larger than the fibre core. This was caused by an f-number (F#) mismatch between the side port tube-lens for the microscope (F# 19) and the MLA (F# 17). Efficiency of light coupling to a fiber optic is highly dependent on the incident angle [34]. Angular mismatch between the MLA (angular variation ± 1.5°) and the 2D fiber optic end (angular variation ± 0.5°) in our system resulted in varied coupling efficiencies between fiber channels. Variation in coupling produces different signal to noise ratio (SNR) for each fiber, which makes our resulting images appear ‘patched’. This becomes more pronounced when imaging relatively dim samples, such as live cells. Further, lower SNR results in larger variation in the estimated fluorescence lifetime, as can be seen for Fluorescein in Fig. 4(b) (square patched) compared to Coumarin-6, which had much higher SNR.
Besides SNR, there are additional factors contributing to the uncertainty in lifetime estimation, e.g., the ratio of observation window to estimated lifetime. A known requirement for accurate estimation of lifetime using MLE is W / τ > 1.5, where τ is the lifetime W is the observation window [35]. The outward sweeping mode decreases crosstalk but it also reduces the observation window to 6.5 ns. Excluding the region at the turnaround points and considering the convolution with the IRF, the observation window containing the exponential decay is only 3.5 ns. Further, our fitting neglects the effect of folding decay at the turnaround points which effects the accuracy of lifetime estimation [36].
Sample photobleaching and the number of photons collected from the sample effect lifetime estimation. The expression of fluorescently labelled proteins in live cells is relatively dim compared to dyes. In addition, these proteins are often localized on membranes and have reduced diffusion rates. In order to collect data from the live cell sample, we extend the acquisition time to acquire sufficient photons. However, extended exposure also increases sample photobleaching, which resulted in the underestimation of mCerulean3 labelled Bcl-XL. In live cells, photobleaching reduces the number of photons collected from the fluorophore-fusion protein of interest while the contribution of auto fluorescence remains unchanged. The ratio of fluorescence signal to auto-fluorescence signal decreases with photobleaching resulting in underestimation of estimated lifetimes. We have replicated this effect on the ISS-Alba microscope, by exposing the cells to higher laser power and slower scan times (data not shown).
It should be noted that, in the case of FLIM, time-correlated single photon counting (TCSPC) is the most widely used technique [37,38]. Comparing to TCSPC, where a single photon is recorded within a single pulse, direct sampling technique such as the streak camera has the advantages of recording single short events and high temporal resolution. However, new multiplexing detector technologies have enabled multiple channel TSCPC [38] and sub-100 ps temporal resolution is sufficient for most FLIM applications.
In conclusion, we have developed a time delay scheme in a multiplexing synchroscan streak camera to minimizing the crosstalk between input channels as well as temporal nonlinearity. We have demonstrated the use of this technique in a FLIM setup to measure spatial lifetime variations in a fixed Convallaria sample as well as on live MCF-7 cells. Overall, we have demonstrated that the combination of our alternating fiber design and temporal corrections pushes the streak camera to be used as a highly multiplexed detector.
Funding
Natural Sciences and Engineering Research Council (NSERC) of Canada (QF: Discovery 365065; QF & DWA: I2IPJ 486815); Canadian Institutes of Health Research (CIHR, QF & DWA: 161934); Ontario Centres of Excellence (90715, QF & DWA).
Acknowledgment
This project is supported in part by the Natural Sciences and Engineering Research Council (NSERC) of Canada (QF: Discovery 365065; QF & DWA: I2IPJ 486815); Canadian Institutes of Health Research (CIHR, QF & DWA: 161934); Ontario Centres of Excellence (90715, QF & DWA).
References
1. C. H. Sarantos and J. E. Heebner, “Solid-state ultrafast all-optical streak camera enabling high-dynamic-range picosecond recording,” Opt. Lett. 35(9), 1389–1391 (2010). [CrossRef] [PubMed]
2. J. Qu, J. Li, and H. Niu, “Fluorescence lifetime imaging microscopy based on a synchroscan streak camera,” Proc. SPIE 5324, 250 (2004). [CrossRef]
3. I. Gorgisyan, R. Ischebeck, C. Erny, A. Dax, L. Patthey, C. Pradervand, L. Sala, C. Milne, H. T. Lemke, C. P. Hauri, T. Katayama, S. Owada, M. Yabashi, T. Togashi, R. Abela, L. Rivkin, and P. Juranić, “THz streak camera method for synchronous arrival time measurement of two-color hard X-ray FEL pulses,” Opt. Express 25(3), 2080–2091 (2017). [CrossRef] [PubMed]
4. E. Floyd, E. T. Gumbrell, J. Fyrth, J. D. Luis, J. W. Skidmore, S. Patankar, S. Giltrap, and R. Smith, “A high spatio-temporal resolution optical pyrometer at the ORION laser facility,” Rev. Sci. Instrum. 87(11), 11E546–1- 11E546–3 (2016).
5. R. J. Smith and T. E. Weber, “A streak camera based fiber optic pulsed polarimetry technique for magnetic sensing to sub-mm resolution,” Rev. Sci. Instrum. 87(11), 11E725–1- 11E725–4 (2016).
6. J. Liang, L. Zhu, and L. V. Wang, “Single-shot real-time femtosecond imaging of temporal focusing,” Light Sci. Appl. 7(1), 42 (2018). [CrossRef] [PubMed]
7. R. V. Krishnan, A. Masuda, V. E. Centonze, and B. Herman, “Quantitative imaging of protein-protein interactions by multiphoton fluorescence lifetime imaging microscopy using a streak camera,” J. Biomed. Opt. 8(3), 362–367 (2003). [CrossRef] [PubMed]
8. J. Campillo and S. L. Shapiro, “Picosecond Streak Camera Fluorometry—A Review,” IEEE J. Quantum Electron. 19(4), 585–603 (1983). [CrossRef]
9. A. Kusumi, A. Tsuji, M. Murata, Y. Sako, A. C. Yoshizawa, S. Kagiwada, T. Hayakawa, S. Ohnishi, and S. Ohnishi, “Development of a Streak-Camera-Based Time-Resolved Microscope Fluorimeter and Its Application to Studies of Membrane Fusion in Single Cells,” Biochemistry 30(26), 6517–6527 (1991). [CrossRef] [PubMed]
10. C. V. Zint, W. Uhring, M. Torregrossa, B. Cunin, and P. Poulet, “Streak camera: a multidetector for diffuse optical tomography,” Appl. Opt. 42(16), 3313–3320 (2003). [CrossRef] [PubMed]
11. G. Satat, B. Heshmat, D. Raviv, and R. Raskar, “All Photons Imaging Through Volumetric Scattering,” Sci. Rep. 6(1), 33946 (2016). [CrossRef] [PubMed]
12. V. Masilamani, B. B. Das, J. Secor, M. AlSalhi, S. Devanesan, S. Prasad, D. Rabah, and R. R. Alfano, “Optical biopsy of benign and malignant tissue by time resolved spectroscopy,” Technol. Cancer Res. Treat. 12(6), 559–563 (2013). [CrossRef] [PubMed]
13. T. Glanzmann, J. P. Ballini, H. van den Bergh, and G. Wagnières, “Time-resolved spectrofluorometer for clinical tissue characterization during endoscopy,” Rev. Sci. Instrum. 70(10), 4067–4077 (1999). [CrossRef]
14. Hamamatsu, “Guide to Streak Cameras,” https://www.hamamatsu.com/resources/pdf/sys/SHSS0006E_STREAK.pdf, (2015).
15. K. Rieger, A. Caldwell, O. Reimann, R. Tarkeshian, and P. Muggli, “GHz modulation detection using a streak camera: Suitability of streak cameras in the AWAKE experiment,” Rev. Sci. Instrum. 88(2), 025110 (2017). [CrossRef] [PubMed]
16. V. K. Ramanujan, J. A. Jo, G. Cantu, and B. A. Herman, “Spatially resolved fluorescence lifetime mapping of enzyme kinetics in living cells,” J. Microsc. 230(Pt 3), 329–338 (2008). [CrossRef] [PubMed]
17. A. Tsikouras, J. Ning, S. Ng, R. Berman, D. W. Andrews, and Q. Fang, “Streak camera crosstalk reduction using a multiple delay optical fiber bundle,” Opt. Lett. 37(2), 250–252 (2012). [CrossRef] [PubMed]
18. Y. Yuan, T. Papaioannou, and Q. Fang, “Single-shot acquisition of time-resolved fluorescence spectra using a multiple delay optical fiber bundle,” Opt. Lett. 33(8), 791–793 (2008). [CrossRef] [PubMed]
19. A. Tsikouras, R. Berman, D. W. Andrews, and Q. Fang, “High-speed multifocal array scanning using refractive window tilting,” Biomed. Opt. Express 6(10), 3737–3747 (2015). [CrossRef] [PubMed]
20. P. Drap and J. Lefèvre, “An exact formula for calculating inverse radial lens distortions,” Sensors (Basel) 16(6), 1–18 (2016). [CrossRef] [PubMed]
21. C. T. Silbernagel, P. Torres III, and D. H. Kalantar, “A Method for Analyzing High Resolution, Time Domain, Streak Camera Calibration Data PULSE,” Proc. SPIE 5559, 435–442 (2004). [CrossRef]
22. M. Liu, M. Jia, H. Pan, L. Li, M. Chang, H. Ren, F. Argoul, S. Zhang, and J. Xu, “Instrument response standard in time-resolved fluorescence spectroscopy at visible wavelength: quenched fluorescein sodium,” Appl. Spectrosc. 68(5), 577–583 (2014). [CrossRef] [PubMed]
23. G. Jones, W. R. Jackson, C. Y. Choi, and W. R. Bergmark, “Solvent effects on emission yield and lifetime for coumarin laser dyes. Requirements for a rotatory decay mechanism,” J. Phys. Chem. 89(2), 294–300 (1985). [CrossRef]
24. M. M. Martin, L. Lindqvist, R. Sjöback, J. Nygren, and M. Kubista, “Absorption and fluorescence properties of fluorescein,” J. Lumin. 51(6), L7–L21 (1995).
25. T. A. Laurence and B. A. Chromy, “Efficient maximum likelihood estimator fitting of histograms,” Nat. Methods 7(5), 338–339 (2010). [CrossRef] [PubMed]
26. P. Poulet, C. V. Zint, M. Torregrossa, W. Uhring, and B. Cunin, “Comparison of two time-resolved detectors for diffuse optical tomography: photomultiplier tube-time-correlated single photon counting and multichannel streak camera,” in Optical Tomography and Spectroscopy of Tissue VB. Chance, R. R. Alfano, B. J. Tromberg, M. Tamura, and E. M. Sevick-Muraca, eds., Proc. SPIE 4955, 154–163 (2003).
27. J. Kale, E. J. Osterlund, and D. W. Andrews, “BCL-2 family proteins: changing partners in the dance towards death,” Cell Death Differ. 25(1), 65–80 (2018). [CrossRef] [PubMed]
28. F. Mérola, A. Fredj, D. B. Betolngar, C. Ziegler, M. Erard, and H. Pasquier, “Newly engineered cyan fluorescent proteins with enhanced performances for live cell FRET imaging,” Biotechnol. J. 9(2), 180–191 (2014). [CrossRef] [PubMed]
29. Q. Liu, B. Leber, and D. W. Andrews, “Interactions of pro-apoptotic BH3 proteins with anti-apoptotic Bcl-2 family proteins measured in live MCF-7 cells using FLIM FRET,” Cell Cycle 11(19), 3536–3542 (2012). [CrossRef] [PubMed]
30. H. Sparks, F. Görlitz, D. J. Kelly, S. C. Warren, P. A. Kellett, E. Garcia, A. K. L. Dymoke-Bradshaw, J. D. Hares, M. A. A. Neil, C. Dunsby, and P. M. W. French, “Characterisation of new gated optical image intensifiers for fluorescence lifetime imaging,” Rev. Sci. Instrum. 88(1), 013707 (2017). [CrossRef] [PubMed]
31. M. Ossiander, F. Siegrist, V. Shirvanyan, R. Pazourek, A. Sommer, T. Latka, A. Guggenmos, S. Nagele, J. Feist, J. Burgdörfer, R. Kienberger, and M. Schultze, “Attosecond correlation dynamics,” Nat. Phys. 13(3), 280–285 (2017). [CrossRef]
32. T. Gaumnitz, A. Jain, Y. Pertot, M. Huppert, I. Jordan, F. Ardana-Lamas, and H. J. Wörner, “Streaking of 43-attosecond soft-X-ray pulses generated by a passively CEP-stable mid-infrared driver,” Opt. Express 25(22), 27506–27518 (2017). [CrossRef] [PubMed]
33. P. Schreiber, S. Kudaev, P. Dannberg, and U. D. Zeitner, “Homogeneous LED-illumination using microlens arrays,” Proc. SPIE 5942(59420K), 59420K (2005). [CrossRef]
34. A. H. Al-hamdani, H. G. Rashid, and Z. R. Ghayib, “Improvement of laser to fiber coupling efficiency,” Arpn 9(11), 2286–2291 (2014).
35. D. D. Li, S. Ameer-Beg, J. Arlt, D. Tyndall, R. Walker, D. R. Matthews, V. Visitkul, J. Richardson, and R. K. Henderson, “Time-domain fluorescence lifetime imaging techniques suitable for solid-state imaging sensor arrays,” Sensors (Basel) 12(5), 5650–5669 (2012). [CrossRef] [PubMed]
36. R. W. K. Leung, S.-C. A. Yeh, and Q. Fang, “Effects of incomplete decay in fluorescence lifetime estimation,” Biomed. Opt. Express 2(9), 2517–2531 (2011). [CrossRef] [PubMed]
37. S. Moon, Y. Won, and D. Y. Kim, “Analog mean-delay method for high-speed fluorescence lifetime measurement,” Opt. Express 17(4), 2834–2849 (2009). [CrossRef] [PubMed]
38. M. Wahl, T. Röhlicke, H. J. Rahn, R. Erdmann, G. Kell, A. Ahlrichs, M. Kernbach, A. W. Schell, and O. Benson, “Integrated multichannel photon timing instrument with very short dead time and high throughput,” Rev. Sci. Instrum. 84(4), 043102 (2013). [CrossRef] [PubMed]
