We compared performance of recently developed silicon photomultipliers (SiPMs) to GaAsP photomultiplier tubes (PMTs) for two-photon imaging of neural activity. Despite higher dark counts, SiPMs match or exceed the signal-to-noise ratio of PMTs at photon rates encountered in typical calcium imaging experiments due to their low pulse height variability. At higher photon rates encountered during high-speed voltage imaging, SiPMs substantially outperform PMTs.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
Two-photon imaging is the leading technique for recording high-resolution fluorescence images in thick or scattering tissue . In two-photon imaging, excitation occurs via nonlinear absorption, and is tightly confined to the focus of a laser beam. Images are formed by scanning the laser through the sample. Because fluorescence is generated at only one point at a time, two-photon microscopes can use a single-pixel detector to record all fluorescence emitted by the sample, regardless of scattering. A variety of single-pixel detectors have been used in two-photon imaging:
Photomultiplier tubes (PMTs) are the most commonly used single-pixel detectors in light microscopy. PMTs detect light at a photocathode, which emits electrons upon photon absorption . Electrons are accelerated under high voltage to strike a series of dynodes , which release additional electrons, exponentially amplifying the signal (Fig. 1(a)). The highest-sensitivity PMTs for visible light use Gallium-Arsenide-Phosphide (GaAsP) photocathodes with high quantum efficiencies (QE; >40%). PMT responses have highly variable amplitudes , because they achieve gain via many sequential low-gain (<10-fold) amplification steps. Stochastic variations from early steps are propagated through subsequent steps, resulting in highly-variable output pulse heights (Fig. 1(b)).
Avalanche photodiodes (APDs) are light-sensitive semiconductor diodes to which a bias voltage is applied. Photon absorption generates electron-hole pairs, which are accelerated by the bias voltage, producing additional electron-hole pairs via impact ionization. At lower bias voltages, APDs operate in linear mode, in which each photon produces a current pulse. At higher bias voltages (“Geiger mode”), runaway ionization produces a saturating signal, causing the APD to act as a photon-gated binary switch. In Geiger mode, additional simultaneously-arriving photons do not produce additional signal. The Geiger mode avalanche is quenched by active or passive circuits , which cause the bias voltage to drop below breakdown then restore it, allowing another photon to be detected. Geiger mode APDs (also called SPADs) have excellent quantum efficiency (>80%) and lower pulse height variability than PMTs, but have small sensitive areas, making single SPADs inefficient for collecting the large etendue of emission light from scattering specimens.
Hybrid detectors (HPDs) combine the large and sensitive photocathodes of PMTs with the single-stage gain of (linear-mode) APDs . HPDs tested in our labs produced large ‘super-pulses’, occurring approximately once per 104 detected photons, caused by high energy photons emitted from the contained avalanche diode striking the photocathode (data not shown). These super-pulses occlude the normal photon signal when they occur, limiting the usefulness of HPDs at higher photon rates.
Silicon photomultipliers (SiPMs) consist of an array of SPADs fabricated on a shared substrate, with outputs into a shared readout channel [7–10] (Fig. 1(c)). Each element acts as an all-or-none photon detector that produces a pulse of stereotypic height with low variability (Fig. 1(d)). Pulses from all elements sum to form the SiPM output. The large number of elements in the array (>1000) allows many photons to be detected simultaneously without saturation, and enable large active areas compatible with large-etendue objectives. However, they also have much larger dark rates (i.e. spurious detections not triggered by photon absorption) than PMTs, and exhibit tradeoffs between element counts, response durations, crosstalk, and QE. SiPMs benefit from low-cost CMOS fabrication, low-voltage operation, long operating life, and insensitivity to magnetic fields. As a result, these detectors have seen over a decade of engineering [11,12] and broad applications in e.g. lidar  and medical imaging .
Arrays of detectors have been used to achieve spectral , spatial , and superresolution  sensitivity in multiphoton imaging, and are particularly useful for increasing the maximum count rate of photon counting techniques [18–20]. Other single-photon detectors [21–23] show desirable characteristics but are currently impractical for microscopy due to e.g. small sensitive areas and cryogenic operating temperatures.
Two-photon imaging experiments historically produced low signal photon rates (<1 MHz), a regime in which PMTs excel due to their low dark rates and high gain. However, advances in reagents and hardware such as brighter fluorescent indicators , excitation and collection optics , and the use of multiple or extended excitation foci [26,27], have progressively increased photon rates in many experiments, enabling faster and more precise measurements. Higher photon rates are particularly important for optical physiology , which involves measuring small fluorescence transients lasting only milliseconds. As photon rates increase, the relative contribution of various noise sources may change, making it useful to reassess the performance of different detectors. In particular, the saturating Geiger-mode mechanism of SiPM amplification results in very low pulse height variability, which is the largest source of noise in other detectors at high photon rates. Unlike PMTs, SiPMs are not damaged by high light levels and saturate slowly, resulting in a wider dynamic range. These factors could make SiPMs better suited for high-speed and high-SNR imaging.
SiPMs have not yet been widely used for two-photon microscopy, perhaps because little information has been available on their performance for biological imaging (but see [27,29,30]). In this study, we compared neural activity recordings using SiPM modules with built-in cooling elements (Multi-Pixel Photon Counters ; MPPCs; Hamamatsu C13366-6067, 14455-8611) to commonly-used GaAsP PMTs selected for high quantum efficiency and low dark rate (Hamamatsu H11706P-40 SEL, H10770PB-40 SEL). The two SiPMs used were an existing blue-sensitive model and a newly developed peak-shifted design with improved sensitivity at green-to-red wavelengths.
We first measured the quantum efficiency, dark rates, and pulse height variability of the SiPMs and a PMT. The SiPMs showed lower quantum efficiency than the PMT, and many-fold higher dark rates (Table 1). However, the distinct gain mechanisms of the detectors resulted in very different pulse height distributions. SiPMs produced a pulse height distribution with clear peaks corresponding to integer numbers of detected fluorescence photons for each laser pulse (Fig. 1(f)), whereas peaks were not distinguishable in the PMT pulse height distribution in matched-intensity recordings (Fig. 1(e)).
To evaluate how these differences in quantum efficiency, dark rate, and pulse height variability affect imaging experiments, we measured the signal-to-noise ratio (mean/standard deviation) of individual pixels in resonant-scanned two-photon recordings of a static fluorescent sample (a pollen grain) across a wide range of pixel intensities, comparing the PMT and blue-sensitive MPPC. We fit a simple model of the sample variance with multiplicative and additive noise sources:1(g)). The SiPM showed higher additive noise than the PMT (αSIPM = 1.26 +- 0.28, αPMT = 0.32 +- 0.29; mean +-95% c.i.; p < 1e-3, two-sample t-test), as expected from its higher dark rate. However, it showed lower multiplicative noise (µSIPM = 2.29 +- 0.41; µPMT = 3.55 +- 0.55; mean +-95% c.i.; p < 1e-3, two-sample t-test) indicating that the SiPM’s low pulse height variability overcomes the effect of its lower quantum efficiency. As a result, the SiPM had lower SNR than the PMT at photon rates below 1 MHz, but higher SNR at photon rates above 1 MHz (Fig. 1 h).
Drosophila melanogaster Kenyon cells are part of the fly olfactory pathway and have large Ca2+ influxes when flies are presented with odors. We performed raster scanning two photon imaging of odor responses in Kenyon cells expressing the genetically encoded Ca2+-indicator GCaMP6f. A 50:50 mirror was used to simultaneously record the sample with the PMT and blue-sensitive SiPM. Images of Kenyon cells (Fig. 2(a)) and odor-evoked calcium transients amplitudes (Fig. 2(b)) captured with the two detectors were qualitatively similar. We quantified the SNR of each cell’s ΔF/F0 response, which was slightly higher in the SiPM recordings (Fig. 2(c), paired two-sided t-test p = 4e-15).
We compared the detectors at higher excitation powers and photon rates encountered in vertebrate imaging by recording from motor cortex of a GCaMP6s-expressing head-fixed awake mouse performing a licking behavior (Fig. 2(d,e)), using a second resonant-scanning microscope with a similar 50:50 mirror detection arrangement and PMT (H10770PB-40 SEL). As with flies, SNR of SiPM recordings was higher than PMT recordings (Fig. 2(f), paired two-sided t-test p = 6e-11);
We next compared the detectors at yet higher photon rates achieved by recently-developed imaging methods. We performed 1 kHz voltage imaging using a SLAP tomographic two-photon microscope, which produces higher photon and frame rates than raster scanning . We imaged cultured rat hippocampal neurons labeled with the red-emitting (∼590-650 nm) voltage-sensitive dye RhoVR1.pip.sulf  while eliciting spiking activity with electrical field stimulation. We compared the blue-sensitive MPPC to the PMT in interleaved trials, and in another set of samples, the blue-sensitive MPPC to the peak-shifted MPPC. We evaluated SNR of each recording as the ratio of evoked spikes to the standard deviation of the pre-stimulation baseline. At the red emission wavelengths of this dye, the peak-shifted MPPC outperformed the blue-sensitive MPPC (SNRred/SNRblue=1.21 +/- 0.04, p = 0.014, paired t-test). The blue-sensitive MPPC, in turn, outperformed the PMT in paired measurements (SNRblue/SNRPMT=1.45 +/- 0.13, p = 0.047, paired t-test). Photon rates in these recordings ranged from 30 to 100 photons per pulse (150-500 MHz).
Our results show that current SiPMs achieve higher SNR than PMTs for activity recordings, particularly high-photon-rate voltage imaging experiments. We expect methods that produce high signal rates, such as second-harmonic or confocal imaging, to also benefit from SiPMs. SNR is only one criterion that should be considered when selecting detectors. SiPMs have practical advantages over PMTs for some applications, and disadvantages for others. PMTs are damaged by high photocurrents, and must be shuttered or gated during light stimulation  experiments. GaAsP photocathode efficiency degrades irreversibly with light exposure. PMTs are also difficult to fabricate and therefore expensive. In contrast, SiPMs are inexpensive CMOS devices, recover completely from saturating light exposure within microseconds, and do not accumulate light damage. However, maximizing SiPM QE requires large active areas per element to minimize dead space, resulting in responses lasting tens to hundreds of nanoseconds (Fig. 3). The SiPMs used here have responses approximately matching the dwell time (∼90 ns) of resonant-scanning systems, and do not cause smearing of images at the pixel spacings recorded here. However, precisely timed signals for e.g. fluorescence lifetime imaging or very rapid scanning [26,27,33] at pixel spacings larger than the microscope’s point spread function would require additional electronic filtering or deconvolution [34,35]. Within a single recording, photon rates vary in space and time, and SiPMs’ additive noise may impair detection of faint events even when SNR is improved overall. Generally, red-sensitive SiPMs have lower quantum efficiencies at a given dark rate and sensor area than blue-sensitive SiPMs. All SiPMs we measured had higher sensitivity than PMTs to the infrared excitation laser (λ∼1um), making it critical to include a short-pass filter (Semrock FF-01-750) in our detection path to avoid laser bleed-through. The cost of a cooled SiPM module and associated electronics is approximately 30% lower than an uncooled PMT.
In conclusion, the SiPMs we tested match or exceed the performance of PMTs in optical physiology experiments, with lower costs, longer lifetime, and simpler operation. When imaging at high frame rates or measuring small fractional changes in intensity, high photon rates are required to detect signals of interest above shot noise, and the dark rates of SiPMs are negligible by comparison. As reagents and imaging conditions continue to improve, photon rates will increase further, and an even greater proportion of experiments will operate in regimes in which SiPMs are superior to PMTs.
4.1 Detector electronics
The PMT used for fly recordings (H11706P-40 SEL) is a gated and shuttered model selected for high quantum efficiency and low dark counts. It was 2 years old, and had not been exposed to extreme light levels. The PMT used for mouse recordings (H10770PB-40 SEL) is a similar shuttered model, with measured QE of 42% and dark count rate of 1.0/ms (compare Table 1). These are representative of high-performance PMTs in regular use in research labs. PMT control voltage was set to 0.63 V for raster recordings. PMT output was amplified using a custom transimpedance amplifier used in other studies [36,37].
We selected MPPCs with a 50 µm pixel pitch for our studies. Larger pixel pitch devices exhibit higher QE, but saturate more quickly at high photon rates and have slower responses incompatible with short pixel dwell times without additional filtering. MPPCs were powered with a linear power supply (+/- 5 V, 0.5 A, Acopian DB5-50), which can supply multiple MPPCs simultaneously. MPPC cooling is integrated and does not require separate power or control circuitry. The MPPC modules used also contain integrated amplifiers, producing transients with a peak voltage of 7.5 mV/photon and a bias of -1.5 V to make use of the full dynamic range of our digitizers (NI PXIe-5170R). For SLAP imaging experiments and QE/dark count recordings, the MPPC module output was recorded directly. For raster imaging experiments, we used a custom 5x, biased post-amplifier to amplify and offset the MPPC outputs to match the -0.5 to 0.5 V range of the digitizers used for the PMT. In all recordings, we made efforts to minimize excess electronic noise, which had amplitude much smaller than the single-photon response for both detectors (Fig. 3). The blue-sensitive and peak-shifted MPPCs used have overvoltages of 3 V and 5 V, respectively. Our analyses (data not shown) indicate that signal-to-noise ratios for two-photon calcium imaging will be further improved by increasing these overvoltages to 5.5-6 V.
4.2 Quantum efficiency, dark rate, and SNR measurements
Voltage traces were recorded at 250 MHz using a National Instruments data acquisition system (NI PXIe-5170R). Dark rate measurements were performed with the SiPMs shuttered. For model C13366-6067, dark rates were measured with 3 different SiPM modules and the mean reported. For the other detectors a single module was measured.
To measure quantum efficiency, a switchable bi-color LED (Dialight 598-8621-207F 525 nm/625 nm emission) was imaged onto the sensitive area of the detector inside a darkened box. Illumination power at each color setting was measured with an integrating sphere (ILX Lightwave OMM-6810B and OMH-6722). Quantum efficiency was then calculated by dividing the measured photon rate on each detector by the measured illumination power at the known LED wavelength. To measure photon rate, the mean single-photon response was first obtained from dark count recordings by dividing the sum baseline-subtracted signal by the number of peaks that exceeded a manually-selected threshold (3 mV) after smoothing with a Gaussian kernel (σ=2 samples). The sum baseline-subtracted signal in the QE recordings was divided by the mean single-photon response, and the dark rate subtracted. This method accounts for cross-talk induced doublet detections  that occur in SiPMs due to semiconductor defects.
In Fig. 1(g), we computed the SNR for Poisson measurements with additive and multiplicative noise, related to Eq(1) by: SNR = x/sqrt(Var(x^)).In Fig. 1(h), we recorded a static sample (a pollen grain) and computed the empirical SNR for each pixel as the mean signal for that pixel divided by its standard deviation in time. We performed least-squares fits of Eq(1) to the pooled pixel data (plotted lines). For clarity, the plotted markers are the mean brightness and SNR for bins of equal pixel counts.
4.3 Pulse height distribution measurements
A SLAP microscope  was used to excite a slice of fixed mouse brain tissue expressing a fluorescent protein. A 50/50 beamsplitter (Chroma 21000) was used to direct emission light simultaneously to the blue-sensitive SiPM and PMT. The digitizer was synchronized to the laser emission via a phase-locked loop. For each detector, signal amplitudes were obtained by linear regression of the detector signal in the 200 ns period following each laser pulse against a template. The template used was the trimmed mean of these 200 ns detector signals, omitting the 20% of responses with largest and smallest peaks. Histograms (Fig. 1(e,f)) depict the distribution of signal amplitudes, scaled to the single-photon amplitude.
4.4 Fly recordings
We used flies whose mushroom body Kenyon cells expressed GCaMP6f (OK107Gal4 x UASGCaMP6f, offspring of Bloomington stock numbers 854 and 52869). Flies were imaged on a Janelia MiMMS 2.0, resonant scanning microscope (janelia.org/open-science/mimms-20-2018) with an Olympus 20x (XLUMPLFLN) objective lens at a frame rate of 30 Hz. A Coherent Ultra II, Ti-sapphire, pulsed laser at 920 nm and 7 mW of average power after the objective was used for excitation. Emitted light was split with a 50:50 beam splitter (Chroma 21000) and collected on a blue-sensitive MPPC and H11706P-40 SEL PMT. Matching band-pass emission filters were used in front of both detectors (Semrock FF01-520/70) and a laser-blocking, wide band-pass filter (Semrock FF01-750) was used before the 50:50 split.
Because the splitting ratio of the 50:50 beamsplitter and emission filters is not necessarily 1:1, a series of measurements were made by swapping the same detector between the two positions while imaging the same sample. In fly recordings, we determined that the SiPM arm of the detection pathway received 1.3x the photons of the PMT arm. We accounted for this in analyses involving the beamsplitter as follows: In Fig. 1(h), the incident photon rate for the SiPM was multiplied by 1.3 for fitting and plotting. In Fig. 2(c), SNR for the SiPM was divided by sqrt(1.3). In mouse recordings, we found that the SiPM arm of the detection path received 0.96x the photons of the PMT arm, and no correction was made for this small difference.
GcaMP6f was expressed in Drosophila Kenyon cells with the genotype OK107Gal4 x UASGCaMP6f (VK5). Flies were mounted and prepared as described previously . Briefly, flies were immobilised with glue and a hole was made in the cuticle above the Kenyon cells. The exposed brain was then covered with 5% agarose (Cambrex Nusieve, #50080) made in bath saline (in mM): NaCl, 103; KCl, 3; CaCl2, 1.5; MgCl2, 4; NaHCO3, 26; N-tris(hydroxymethyl) methyl-2-aminoethane- sulfonic acid, 5; NaH2PO4, 1; trehalose, 10; glucose, 10 (pH 7.3 when bubbled with 95% O2 and 5% CO2, 275 mOsm).
4.5 Mouse recordings
We recorded from a single CamK2a-tTA (JAX:007004) x Ai94(TITL-GCaMP6s)39 (JAX: 024104) crossed with Emx1-Cre (JAX: 005628) mouse. All surgical procedures were in accordance with protocols approved by the HHMI Janelia Research Campus Institutional Animal Care and Use Committee (IACUC 17–155). Surgical procedures were performed under 1-2% isoflurane anaesthesia. Buprenorphine HCl (0.1 ml, 0.03 mg/ml) was administered after surgery. Ketoprofen (0.1 ml, 0.1 mg/ml) was provided for three days following surgery. A 3 mm circular craniotomy was centered over ALM (2.5 mm anterior and 1.5 mm lateral from Bregma). The craniotomy was covered by a cranial window composed of 3 layers of circular glass (total thickness, 450 mm), consisting of two 2.5 mm diameter disks and a 3.5 mm disk that rested on the skull. The window was cemented in place using cyanoacrylate glue and dental acrylic (Lang Dental). A custom headbar was attached just anterior to the window using cyanoacrylate glue and dental cement. During imaging, the mouse was headfixed on the microscope and provided occasional water via a remotely-controlled lickport to encourage licking. Imaging was performed using a customized Janelia MiMMS microscope similar to that used for fly experiments.
4.6 Analysis of in vivo recordings
Regions of interest (ROIs) were manually drawn around cells in the field of view, and the average pixel intensities within each ROI extracted from acquired image time series files. The ΔF/F0 value for each cell at each time-point was computed as ΔF/F0 = (Fi - F0)/F0, where Fi is the raw intensity of a ROI at a given time-point and F0 was the minimum intensity of the ROI within the recording smoothed with a 0.5s moving average filter. Estimated SNR was computed by fitting a non-negative AR(1) process (τdecay=1 second) using non-negative deconvolution  to the ΔF/F0 traces, and taking the ratio of the root-mean-square fit value to the root-mean-square residual.
4.7 Voltage imaging
Hippocampal neuron cultures and buffers were prepared as described previously . After 7 days in vitro, neurons were loaded with 500nM RhoVR1.pip.sulf for 20 minutes, washed in imaging buffer, and imaged at room temperature. We applied 10 field stimuli (90 V/cm) at 20 Hz in the presence of synaptic blocking drugs (10 µM CNQX, 10 µM CPP, 10 µM gabazine, 1 mM MCPG) to elicit single action potentials per stimulus.
For any given field of view, a neuron soma was selected for imaging using the SLAP microscope, and remaining pixels were blocked by the microscope’s spatial light modulator. All collected light from the SLAP microscope was directed to one of two detection arms on each trial, through the same 575-725 nm filter (Semrock FF01-650/150). We performed pairwise comparisons between the blue-sensitive SiPM and the PMT, and between the blue-sensitive and peak-shifted SiPMs. When using the PMT, control voltage was set to 0.75 V, approximately the maximum that did not result in overcurrent protections triggering during our experiments. The PMT signal was conditioned using a custom 10x post-amplifier (40 MHz bandwidth) to match the signal to the input range and broaden pulses to ensure adequate oversampling by the 250 MHz SLAP microscope digitizer. For each neuron, a single trial (1.4s) was collected on one detector, then a dichroic was added/removed from the detection path, sending light to the other detector for the next trial. The detector used first was switched for each field of view to avoid any ordering effects.
We computed the SNR of each recording as the mean amplitude of the 10 evoked spikes divided by the standard deviation of the 500 ms period preceding stimulation, after removing low frequencies (<10 Hz) to compensate for bleaching or photobrightening.
4.8 Software and design document availability
Analysis was performed in Matlab. Analysis and plotting scripts are available upon request. Mechanical and electronic designs, e.g. for detector mounts and post-amplifiers used in this study, are available via the Janelia Open Science Portal for the SLAP microscope: janelia.org/open-science/kilohertz-frame-rate-tomographic-2-photon-microscope/
Howard Hughes Medical Institute.
We thank Deepika Walpita and Steven Sawtelle for experimental support, and Spencer LaVere Smith for comments on the manuscript. We thank Hamamatsu Corporation for customizing MPPC module electronics to our specifications and making early production versions available to us for purchase. We thank Dino Butron, Javier Jurado, Jake Li, Tsuyoshi Ota, Adam Palmentieri, Kathryn Pritchard, and Kotaro Ujihara of Hamamatsu for information on MPPCs and comments on the manuscript.
The authors have no competing financial interests in this study.
1. W. R. Zipfel, R. M. Williams, and W. W. Webb, “Nonlinear magic: multiphoton microscopy in the biosciences,” Nat. Biotechnol. 21(11), 1369–1377 (2003). [CrossRef]
2. V. K. Zworykin, G. A. Morton, and L. Malter, “The Secondary Emission Multiplier-A New Electronic Device,” Proc. IRE 24(3), 351–375 (1936). [CrossRef]
3. J. Rajchman and E. W. Pike, Electrostatic Focusing in Secondary Emission Multipliers. (RCA Manufacturing Co. RCA Victor Division Electronic Research Laboratory, 1937).
4. J. D. Driscoll, A. Y. Shih, S. Iyengar, J. J. Field, G. A. White, J. A. Squier, G. Cauwenberghs, and D. Kleinfeld, “Photon counting, censor corrections, and lifetime imaging for improved detection in two-photon microscopy,” J. Neurophysiol. 105(6), 3106–3113 (2011). [CrossRef]
5. S. Cova, M. Ghioni, A. Lacaita, C. Samori, and F. Zappa, “Avalanche photodiodes and quenching circuits for single-photon detection,” Appl. Opt. 35(12), 1956–1976 (1996). [CrossRef]
6. E. Albrecht, M. Alemi, G. Barber, J. Bibby, M. Campbell, A. Duane, T. Gys, J. Montenegro, D. Piedigrossi, R. Schomaker, W. Snoeys, S. Wotton, and K. Wyllie, “Performance of hybrid photon detector prototypes with 80% active area for the rich counters of LHCB,” Nucl. Instrum. Methods Phys. Res., Sect. A 442(1-3), 164–170 (2000). [CrossRef]
7. S. Tisa, A. Tosi, and F. Zappa, “Fully-integrated CMOS single photon counter,” Opt. Express 15(6), 2873–2887 (2007). [CrossRef]
8. P. Buzhan, B. Dolgoshein, L. Filatov, A. Ilyin, V. Kantzerov, V. Kaplin, A. Karakash, F. Kayumov, S. Klemin, E. Popova, and S. Smirnov, “Silicon photomultiplier and its possible applications,” Nucl. Instrum. Methods Phys. Res., Sect. A 504(1-3), 48–52 (2003). [CrossRef]
9. E. Grigoriev, A. Akindinov, M. Breitenmoser, S. Buono, E. Charbon, C. Niclass, I. Desforges, and R. Rocca, “Silicon photomultipliers and their bio-medical applications,” Nucl. Instrum. Methods Phys. Res., Sect. A 571(1-2), 130–133 (2007). [CrossRef]
10. K. Yamamoto, K. Yamamura, K. Sato, T. Ota, H. Suzuki, and S. Ohsuka, “Development of Multi-Pixel Photon Counter (MPPC),” in 2006 IEEE Nuclear Science Symposium Conference Record (2006), 2, pp. 1094–1097.
11. N. Dinu, R. Battiston, M. Boscardin, G. Collazuol, F. Corsi, G. F. Dalla Betta, A. Del Guerra, G. Llosá, M. Ionica, G. Levi, S. Marcatili, C. Marzocca, C. Piemonte, G. Pignatel, A. Pozza, L. Quadrani, C. Sbarra, and N. Zorzi, “Development of the first prototypes of Silicon PhotoMultiplier (SiPM) at ITC-irst,” Nucl. Instrum. Methods Phys. Res., Sect. A 572(1), 422–426 (2007). [CrossRef]
12. Z. Sadygov, A. Olshevski, I. Chirikov, I. Zheleznykh, and A. Novikov, “Three advanced designs of micro-pixel avalanche photodiodes: Their present status, maximum possibilities and limitations,” Nucl. Instrum. Methods Phys. Res., Sect. A 567(1), 70–73 (2006). [CrossRef]
13. R. Agishev, A. Comerón, J. Bach, A. Rodriguez, M. Sicard, J. Riu, and S. Royo, “Lidar with SiPM: Some capabilities and limitations in real environment,” Opt. Laser Technol. 49, 86–90 (2013). [CrossRef]
14. K. Wagatsuma, K. Miwa, M. Sakata, K. Oda, H. Ono, M. Kameyama, J. Toyohara, and K. Ishii, “Comparison between new-generation SiPM-based and conventional PMT-based TOF-PET/CT,” Physica Medica 42, 203–210 (2017). [CrossRef]
15. D. K. Bird, K. W. Eliceiri, C.-H. Fan, and J. G. White, “Simultaneous two-photon spectral and lifetime fluorescence microscopy,” Appl. Opt. 43(27), 5173–5182 (2004). [CrossRef]
16. J. W. Cha, E. Y. S. Yew, D. Kim, J. Subramanian, E. Nedivi, and P. T. C. So, “Non-descanned multifocal multiphoton microscopy with a multianode photomultiplier tube,” AIP Adv. 5(8), 084802 (2015). [CrossRef]
17. M. Castello, G. Tortarolo, M. Buttafava, T. Deguchi, F. Villa, S. Koho, L. Pesce, M. Oneto, S. Pelicci, L. Lanzanó, P. Bianchini, C. J. R. Sheppard, A. Diaspro, A. Tosi, and G. Vicidomini, “A robust and versatile platform for image scanning microscopy enabling super-resolution FLIM,” Nat. Methods 16(2), 175–178 (2019). [CrossRef]
18. S. A. Castelletto, I. P. Degiovanni, V. Schettini, and A. L. Migdall, “Reduced deadtime and higher rate photon-counting detection using a multiplexed detector array,” J. Mod. Opt. 54(2-3), 337–352 (2007). [CrossRef]
19. D. Gudkov, G. Gudkov, B. Gorbovitski, and V. Gorfinkel, “Enhancing the Linear Dynamic Range in Multi-Channel Single Photon Detector Beyond 7OD,” IEEE Sens. J. 15(12), 7081–7086 (2015). [CrossRef]
20. A. Fukasawa, Y. Egawa, T. Ishizu, A. Kageyama, A. Kamiya, T. Muramatsu, G. Nakano, and Y. Negi, “Multichannel HPD for high-speed single photon counting,” Nucl. Instrum. Methods Phys. Res., Sect. A 812, 81–85 (2016). [CrossRef]
21. A. Bross, E. Flattum, D. Lincoln, S. Grünendahl, J. Warchol, M. Wayne, and P. Padley, “Characterization and performance of visible light photon counters (VLPCs) for the upgraded DØ detector at the Fermilab Tevatron,” Nucl. Instrum. Methods Phys. Res., Sect. A 477(1-3), 172–178 (2002). [CrossRef]
22. F. Mattioli, Z. Zhou, A. Gaggero, R. Gaudio, R. Leoni, and A. Fiore, “Photon-counting and analog operation of a 24-pixel photon number resolving detector based on superconducting nanowires,” Opt. Express 24(8), 9067–9076 (2016). [CrossRef]
23. B. Cabrera, R. M. Clarke, P. Colling, A. J. Miller, S. Nam, and R. W. Romani, “Detection of single infrared, optical, and ultraviolet photons using superconducting transition edge sensors,” Appl. Phys. Lett. 73(6), 735–737 (1998). [CrossRef]
24. H. Dana, Y. Sun, B. Mohar, B. K. Hulse, A. M. Kerlin, J. P. Hasseman, G. Tsegaye, A. Tsang, A. Wong, R. Patel, J. J. Macklin, Y. Chen, A. Konnerth, V. Jayaraman, L. L. Looger, E. R. Schreiter, K. Svoboda, and D. S. Kim, “High-performance calcium sensors for imaging activity in neuronal populations and microcompartments,” Nat. Methods 16(7), 649–657 (2019). [CrossRef]
25. A. Singh, J. D. McMullen, E. A. Doris, and W. R. Zipfel, “Comparison of objective lenses for multiphoton microscopy in turbid samples,” Biomed. Opt. Express 6(8), 3113–3127 (2015). [CrossRef]
26. R. Prevedel, A. J. Verhoef, A. J. Pernia-Andrade, S. Weisenburger, B. S. Huang, T. Nöbauer, A. Fernández, J. E. Delcour, P. Golshani, A. Baltuska, and A. Vaziri, “Fast volumetric calcium imaging across multiple cortical layers using sculpted light,” Nat. Methods 13(12), 1021–1028 (2016). [CrossRef]
27. A. Kazemipour, O. Novak, D. Flickinger, J. S. Marvin, A. S. Abdelfattah, J. King, P. M. Borden, J. J. Kim, S. H. Al-Abdullatif, P. E. Deal, E. W. Miller, E. R. Schreiter, S. Druckmann, K. Svoboda, L. L. Looger, and K. Podgorski, “Kilohertz frame-rate two-photon tomography,” Nat. Methods 16(8), 778–786 (2019). [CrossRef]
28. V. Emiliani, A. E. Cohen, K. Deisseroth, and M. Häusser, “All-Optical Interrogation of Neural Circuits,” J. Neurosci. 35(41), 13917–13926 (2015). [CrossRef]
29. R. Liu, Z. Li, J. S. Marvin, and D. Kleinfeld, “Direct wavefront sensing enables functional imaging of infragranular axons and spines,” Nat. Methods 16(7), 615–618 (2019). [CrossRef]
30. M. G. Giacomelli, “Evaluation of silicon photomultipliers for multiphoton and laser scanning microscopy,” J. Biomed. Opt. 24(10), 106503 (2019). [CrossRef]
31. P. E. Deal, R. U. Kulkarni, S. H. Al-Abdullatif, and E. W. Miller, “Isomerically Pure Tetramethylrhodamine Voltage Reporters,” J. Am. Chem. Soc. 138(29), 9085–9088 (2016). [CrossRef]
32. E. A. Naumann, A. R. Kampff, D. A. Prober, A. F. Schier, and F. Engert, “Monitoring neural activity with bioluminescence during natural behavior,” Nat. Neurosci. 13(4), 513–520 (2010). [CrossRef]
33. J. Wu, A. H. L. Tang, A. T. Y. Mok, W. Yan, G. C. F. Chan, K. K. Y. Wong, and K. K. Tsia, “Multi-MHz laser-scanning single-cell fluorescence microscopy by spatiotemporally encoded virtual source array,” Biomed. Opt. Express 8(9), 4160–4171 (2017). [CrossRef]
34. M. Akiba, K. Inagaki, and K. Tsujino, “Photon number resolving SiPM detector with 1 GHz count rate,” Opt. Express 20(3), 2779–2788 (2012). [CrossRef]
35. K. Podgorski and K. Haas, “Fast non-negative temporal deconvolution for laser scanning microscopy,” J. Biophotonics 6(2), 153–162 (2013). [CrossRef]
36. S. S. Kim, H. Rouault, S. Druckmann, and V. Jayaraman, “Ring attractor dynamics in the Drosophila central brain,” Science 356(6340), 849–853 (2017). [CrossRef]
37. N. Li, K. Daie, K. Svoboda, and S. Druckmann, “Robust neuronal dynamics in premotor cortex during motor planning,” Nature 532(7600), 459–464 (2016). [CrossRef]
38. S. Vinogradov, “Analytical models of probability distribution and excess noise factor of solid state photomultiplier signals with crosstalk,” Nucl. Instrum. Methods Phys. Res., Sect. A 695, 247–251 (2012). [CrossRef]
39. T. Hige, Y. Aso, M. N. Modi, G. M. Rubin, and G. C. Turner, “Heterosynaptic Plasticity Underlies Aversive Olfactory Learning in Drosophila,” Neuron 88(5), 985–998 (2015). [CrossRef]