Genetic tools and especially genetically encoded fluorescent reporters have given a special place to optical microscopy in drosophila neurobiology research. In order to monitor neural networks activity, high speed and sensitive techniques, with high spatial resolution are required. Structured illumination microscopies are wide-field approaches with optical sectioning ability. Despite the large progress made with the introduction of the HiLo principle, they did not meet the criteria of speed and/or spatial resolution for drosophila brain imaging. We report on a new implementation that took advantage of micromirror matrix technology to structure the illumination. Thus, we showed that the developed instrument exhibits a spatial resolution close to that of confocal microscopy but it can record physiological responses with a speed improved by more than an order a magnitude.
© 2014 Optical Society of America
Recent developments of fluorescent proteic sensors and optogenetic tools have given a particular place to optical microscopy for in vivo investigation of biological molecular processes. Owing to powerful genetic tools [1, 2] and easy transgenesis , drosophila neurobiology has taken full advantages of these developments [4–8]. In this field, current optical implementations rely on commercial optical set-up, usually either conventional wide-field epifluorescence, confocal or two-photon microscopes. Integrated analysis of most brain functions, like olfaction or memory for instance, would benefit from a global 3D monitoring of activity of the neurons involved in the function. This is out of reach of current approaches. Wide-field epifluorescence microscopy provides a high speed imaging  but good spatial resolution is obtained only for thin samples. The drosophila brain contains about 105 neurons and its dimensions are about 1 mm×0.3 mm×0.2 mm. In the context of drosophila brain studies, as for other thick samples, the lack of optical sectioning capability makes quantification difficult and the spatial resolution of images is poor. On the other hand, confocal or two-photon microscopy can realize high resolution optical sections but they scan sequentially each pixel. Even if high speed scanners have been developed [10, 11], the limited brightness of the biological samples sets a minimum pixel dwelling time and it makes the approach slow. Random-access imaging , which takes advantage of acousto-optical deflector technology to ”jump” sequentially through a limited set of points of interest overcomes the speed limitation, but for living animal imaging, residual movements are a serious issue. In fact, speed requires to increase the parallelism of pixel acquisition. An intermediate trade-off is multifocal point scanning . While also developed for two-photon microscopes [14–16], the fastest and the most mature technology is spinning-disk confocal microscopy implemented with a microlenses array . This technique is convenient for a wide range of applications but it has two drawbacks for our in vivo imaging in neuro-imaging. First, it lacks flexibility. Pinholes size cannot be changed and either the microscope works at the optimal spatial resolution (usual solution in commercial versions) and the sensitivity is poor as soon as we go deeper in the tissue (typical diffusion length in the brain is around 50 μm) or it always works with a reduced spatial resolution. In addition, the field of view is fixed and there is no control of the size of the irradiated area. Secondly, its cost is quite high. Simpler and as efficient approaches can be obtained by going further back to wide-field microscopy but adding sectioning capability. Currently, one of the most successful approaches in developmental biology is selective plane illumination microscopy (SPIM) . Unfortunately, Drosophila cuticle has a bad transmission and for imaging a small part of the cuticle is removed at the top of the head to get an optical access to the brain. This geometry is not compatible with SPIM implementation. T. Wilson’s group had introduced an alternative technique for optical sectioning in wide-field microscopy  based on structured illumination. While the technique generates more photobleaching and cannot handle too dense labeling since it is based on an image subtraction principle, its implementation is more straightforward. In its initial version, the technique was quite slow (an acquisition rate of a few Hz when optimized) and suffered from several artefacts (see ). By introducing the HiLo principle [21, 22], Mertz’s group largely improved the robustness of structured illumination approach against artefact and already improved speed in last implementation, with acquisition rate around 10 Hz . This is still rather slow and in addition, this in vivo implementation was neither optimized nor characterized for high spatial resolution imaging. In this paper, we present a new implementation. It relies on a structured illumination strategy based on the HiLo concept but we took advantage of the micro-mirror array technology (DLP) to structure the illumination. The benefit is a tenfold increase of speed, reaching acquisition rate necessary for 3D monitoring of specific neural networks. We also took advantage of new high power LED illumination instead of laser illumination which avoids a trade-off between image quantification accuracy and speed, since averaging is necessary with laser illumination to remove granularities introduced by random nature of speckle patterns. We showed that this new implementation was compatible with high resolution imaging, with performance close to confocal microscopy but with acquisition rates above video rate.
2. Experimental set-up
The set-up (Fig. 1) is close to a conventional epifluorescence microscope for GFP protein fluorescence imaging. The main difference is the matrix of micro-mirrors that shapes the illumination. An incoherent light source, the ultra-high power LED Prizmatix, provides the excitation beam at 460 nm with a spectral width of 27 nm. The beam is filtered by a bandpass filter at 472 ± 15nm and is sent to a matrix of rotatable micro-mirrors (DLP Texas Instruments Discovery 4100 0.7 XGA). This micro-electro-mechanical spatial light modulator is used to shape the excitation beam. Then, the beam is collimated before being reflected by a dichroic beamsplitter and sent onto the back aperture of a water immersion objective (Leica 40× 0.8 NA or Leica 25× 0.95 NA). The focal plane of the objective is conjugated with the DLP. The fluorescence emission from the sample goes through the objective. The output collimated beam is filtered by a emission filter at 525 nm with a spectral width of 35 nm and is imaged thanks to the tube lens onto a sCMOS camera (model Neo, Andor). The whole experiment is synchronized by a Lab-view program that we developed on a national instrument plateform, through a multifunction data acquisition NI-USB bus to control accurately time during the image acquisition.
In the original implementation of structured illumination microscopy by Mertz’s group, mechanical displacements of a diffuser plate  or a moving grid  limit the acquisition speed to video rate at the most after a tricky optimization. In order to shape the laser beam and generate different illumination patterns, we use the DLP high speed module developed by Vialux which is completely configured by a high speed FPGA logic and a USB controller firmware. Each micro-mirror of the DLP can move independently in two positions, either the reflection goes to the sample (the ”on” position) or not (the ”off” position). After optimization of the Labview control program, we reached a 2 kHz switching rate which corresponds to a 500 μs illumination period with a DLP switching time of 6 μs. It’s therefore not the pattern switch which limits the acquisition speed but only the fluorescence amount emitted by the sample and the signal-to-noise ratio. In addition, thanks to the micromirror device, the pattern can be changed in real time with imaging depth to keep the best compromise between resolution and brightness. As the light goes deep in the tissue, aberrations and scattering decrease the transmission of high frequency components. So it is crucial to be able to reduce the frequency of the illumination pattern to still get a signal even if the axial resolution is degraded. In fact, decreasing the frequency of the grid illumination corresponds in term of confocal microscopy to use a pinhole larger and so to get a thicker optical section. In a confocal microscope the pinhole size is however usually fixed during the recording of a 3D stack of images. Thanks to the high flexibility of the DLP, it is possible to adapt the pattern during the recording.
In the initial implementation, that was used for some of the characterization experiments, the excitation light source was a 488 nm diode laser with a maximum power of 50 mW. While it provided more power at the level of the sample than the LED, when we imaged biological samples with fine details, we realized that the captured section was degraded by random speckles of the illumination. Indeed, speckles were no longer averaged when switching from the slow moving diffuser to the fast switching micro-mirror matrix. Then, image details lose contrast and image quantification can be compromised if internal fluctuations of the living sample make the speckle pattern change. To conciliate high-speed imaging and resolution, we replace the laser by an ultra-high power LED.
3. Image processing
Image processing takes advantage of the robust HiLo approach introduced by J. Mertz’s group . An optical section requires to record two images, one with an uniform illumination and one with an illumination structured by a pattern that contains only high spatial frequencies except for a continuous component (high frequency grid pattern in our set-up). From those two images, high and low spatial frequency components of an optical section around the in-focus plane will be extracted separately (out of focus light contribution is suppressed for both types of components) and then recombined to get the final section. Properties of the optical transfer function of the microscope already ensure that high frequency components of the uniformly illuminated image come specifically from the in-focus plane, so high frequency components of the section Ihi(ρ⃗) are extracted straightforwardly by high-pass filtering of the uniform image. The difficulty is the extraction of low frequency components. This extraction relies on properties of the image acquired with structured illumination. In our application, we aimed at high spatial resolution imaging in thick tissue and the rejection of out of focus light is more critical than for initial demonstration of the technique. So the initial algorithm of Mertz’s group was slightly modified to improve this rejection. In more detail, the image intensities are given in Eq. (1):Eq. (2), that reads: Eq. (2)), the low frequency components of the section Iin(ρ⃗) are not contained in the low frequency region but around the frequency of the grid in the frequency plane of the difference image. It is an optical analog of the radio amplitude modulation (AM). The modulation of the illumination acts as a high frequency ”carrier” for the information of the sectioned image. To specifically extract it and to remove as much as possible the noise, it is helpful to filter the difference image to keep spectral components close to the grid frequency (with a bandwidth corresponding to that of the information to be extracted). Thus, we applied a filter F2g composed of two gaussian functions centered on the illumination pattern spatial frequencies ±k⃗g. It is the main difference with J. Mertz’s group treatment . The expression of the filter F2g reads:
The σ width parameter is chosen to optimize the bandpass around the pattern frequency to extract correctly the in-focus low frequency information. The remaining of the treatment to extract the low frequency information is then the same as previous treatment. It is basically a conventional demodulation, as done to extract the low frequency audio signal from the AM radio signal with a low-pass filter complementary to the high-pass filter used for Ihigh(ρ⃗) extraction. Thus, we got Ilow(ρ⃗), the low frequency components of Iin(ρ⃗). The optical section is finally got by combining the image Ihigh(ρ⃗) and Ilow(ρ⃗) containing respectively high and low frequency components of the section with a reconstruction factor η as given in Eq. (4). This factor equals π if the fundamental frequency of the grid pattern is well transmitted and it is inversely proportional to the modulation of the projected pattern. In complex samples that affects the OTF and η can be determined by matching the spectral densities in the images Ihigh(ρ⃗) and Ilow(ρ⃗) at the cutting frequency of the complementary high and low-pass filters used for Ihigh(ρ⃗) and Ilow(ρ⃗) extractions.
4. Characterization of the microscope
First, we studied the Point Spread Function (PSF) of our microscope. We imaged 100 nm beads doped with fluorescein in agarose, a diameter well below the resolution of our system. For such a point object, there is no difference between the HiLo image and the wide-field image except the removal of out-of-focus wings for defocused images, that are already quite small. Thus, the determination of lateral and axial resolutions was carried out only on uniform images. We got mean lateral and axial bead sizes of 440 ± 40 nm and 2.0 ± 0.1 μm respectively, assuming gaussian fits  (see Fig. 2). This led to a lateral PSF of 430 ± 40 nm and an axial PSF of 2.0 ± 0.1 μm thanks to Eq. (5) that reads:
Assuming a diffraction-limited system , theoretical lateral and axial Full Width Half Maximum (FWHM) of the PSF (see Eq. (6)) equal to 340 nm and 1.8 μm respectively for a numerical aperture NA of 0.8, an index of refraction in the medium between the focal point and the objective of 1.33 and an wavelength λ of 520 nm (peak emission of fluorescein). Comparing experimental and theoretical lateral PSF, the effective numerical aperture NAeff of the objective equals to 0.6.
We completed the characterization of our microscope by determining its spatial frequency response. We imaged DLP patterns onto a plane mirror as sample. The DLP was illuminated with a white light source filtered by a large band blue filter (Schott, BG12). We tested different line periods from 2 to 64 DLP pixels for the illumination pattern shaped by the DLP. Thus, contrary to the previous experiment, the frequency response depends on the illumination arm. A line period of 2 DLP pixels is the smallest that we can create due to the DLP and one DLP pixel is a square of 13.7 μm side. We determined the contrast C of the lines by calculating the standard deviation of the difference of the two normalized images (with uniform and structured illumination) where only the first harmonic of the projected pattern is kept as given in Eq. (7).
σ refer to standard deviation. Results of the contrast measurements are given in Fig. 3. The contrast is maximal and equals to 1 for the 64 DLP pixels lines and decreases when the line spatial frequency increases. The cut-off frequency of our optical system seems to be about 50 ± 10 lines.mm−1. It corresponds to a size of 500 ± 100 nm for the smallest object resolvable by the optical system. This is in good agreement with the lateral PSF determined by the previous experiment. The theoretical optical transfer function (OTF) of an incoherent diffraction-limited system with circular pupils  is given by the Eq. (8) and also represented in Fig. 3.Eq. (9) where m is the 40 magnifying factor of the objective. We got for a half-contrast a frequency of about 18 lines.mm−1 and 15 lines.mm−1 experimentally and theoretically respectively. We observed an experimental constrast close but slightly above the theoretical one. This is related to terms that we have not taken into account in our simple theoretical model. First, even in the absence of any true modulation of the recorded image in structured illumination, the photon noise that is only partially eliminated through our bandwidth filter, will give a small contribution to the contrast. In addition, the projected pattern is a grid and for low frequency modulation several harmonics with a large amplitude are transmitted by the optical system. We applied a gaussian filter to eliminate most of them but for low frequency modulation they give also a small contribution.
We evaluated experimentally the optical sectioning strength of our microscope with a thin fluorescent sample. We spin-coated a poly(methyl methacrylate) film (1% weight vs volume) doped with Rhodamine 6G on a glass slide. We focused on the evolution of the axial resolution for three different line periods: 4, 8, and 16 DLP pixels corresponding to patterns of 1, 2 and 4 μm period respectively on the sample (Fig. 4). The Hilo intensity is maximal when the focal plane of the objective coincides with the thin fluorescent sample and decreases gradually to approach zero. The thickness of the HiLo optical section increases with the fringe period of the pattern. We got for a gaussian fit of the Hilo intensity a FWHM of 5.5 μm, 3.3 μm and 2.0 μm for 16, 8 and 4 DLP pixels line periods respectively. Starting from analytical expression of transfer function of perfect imaging system in incoherent light , we can derive the expression of the sectioning thickness when only the first harmonic of the grating is kept to reconstruct the section. Its expression is given in Eq. (10) that reads:Fig. 3, the line contrast is degraded for lines with frequency near the cut-off frequency of the optical system. The effective number of photons from the section contributing to the final image decreases leading to a degradation of the signal-to-noise ratio. A compromise between axial resolution and sizeable contrast has to be found to determine the fitted line periods. 8 (and eventually 4) DLP pixels line period is suitable to our biological applications.
Finally, to evaluate the quality of the reconstruction from a stack of sections, we imaged 2 μm fluorescent beads. We plotted the intensity cross-section along the spatial dimensions. As expected, we got 2.2 ±0.1 μm for both transversal dimensions of the bead in accord with the lateral PSF (Fig. 5(a)). In order to evaluate the axial resolution, we follow the bead center pixel intensity all along the HiLo images of the 3D stack (Fig. 5(b)). On the Fig. 6, results of axial resolution for wide-field and HiLo microscopy are shown. As the bead size is well above diffraction limit, we observed clearly out-of-focus light rejection for HiLo images. Gaussian fits gave a 2.8 μm FWHM for the HiLo image compared to 5.5 μm for the wide field image, where there is no efficient rejection of the out-of-focus light. Fluorescent beads are well resolved in the three dimensions. Thus, our microscope is suitable to our biological samples with 2 μm typical size structures.
5. Application to biology: Morphological and functional studies of drosophila brain
We evaluated our microscope’s performance on model samples as thin fluorescent sample and known diameter beads. We checked the ability to find this performance when we imaged a real biological sample. Here we present results of in vivo imaging of the drosophila melanogaster brain and especially of a region called the mushroom bodies (MB) . The MB were demonstrated to be the center of the olfactive memory system [8, 30]. Flies carrying the genetically encoded fluorescent probes UAS-CD8-GFP or UAS-NLS-GFP (NLS is set for Nuclear Localization Signal) were crossed with 238Y-Gal4 flies to drive GFP expression in all MB intrinsic neurons, so-called Kenyon cells, thanks to the UAS-Gal4 system . The flies were reared at a controlled temperature (25°C). For in vivo imaging, the fly was glued by the dorsal part of its head and thorax on a plastic film. Then a small aperture was made in the plastic film at the level of the head to remove the underlying cuticle and trachae. The brain was bathed during the experiment with physiological Ringer’s solution containing (in mM) 130 NaCl, 5 KCl, 36 C12H22O11 sucrose, 2mM MgCl2, 2mM CaCl2 and 5 HEPES NaOH [pH = 7.3] (see  for more details about fly preparation).
We compared images taken by our microscope with those taken by confocal microscopy, a high resolution reference technique for drosophila neuro-imaging. Figure 7 shows images of the soma of Kenyon cells for flies of genotype 238Y-Gal4/+; UAS-NLS-GFP/+. These flies expressed a nuclear GFP specifically in MB Kenyon cell nucleus. The Fig. 7(a) obtained with uniform illumination corresponds to traditional wide-field microscopy. The absence of sectioning ability of this technique resulted in a blurred image. Structured illumination image is shown on Fig. 7(b). It was also a blurred image with a strip pattern but, combined with the uniform illumination image (Fig. 7(a)), it can be processed with the numerical treatment described above to reconstruct the section shown on Fig. 8(a). Out-of-focus fluorescence was rejected. The genetically encoded fluorescent probe UAS-GFP.NLS is expressed in the cell nucleus of about 2 μm typical size, which are well resolved by our microscope which gave equivalent images to those of the confocal microscope (Fig. 8(b)). In Fig. 9, HiLo and confocal images were obtained by crossing 238Y-Gal4 flies and UAS-CD8-GFP flies so that GFP was targeted to plasma membranes of Kenyon cells. It is noteworthy that the images of the cell bodies nucleus and the cell bodies membranes looks like complementary even though the typical size of a these neuron cell bodies are closed to 2 μm. Our microscope kept an excellent resolution even for in vivo imaging where the main limitations for spatial resolution came from sample aberrations and diffusion and to a smaller extend from residual small movements of the preparation. Therefore, our micromirror structured illumination microscope working with an incoherent excitation source succeeded in high resolution imaging as the commonly used confocal microscope.
The scanning time of the confocal microscopy is a major limiting factor for in vivo imaging especially to follow in real time neurobiological functional responses. A last experiment was realized to demonstrate the main advantage of our set-up over confocal : sensitivity and speed. We followed the response to electric shock stimuli by fast calcium imaging. The stimulus was delivered in constant current mode through two gold electrodes touching the fly’s body. The flies were obtained by crossing flies carrying the genetically encoded calcium sensor Gcamp3 with 238Y-Gal4 flies. Gcamp3 fluorescence rate increases with the calcium concentration under a constant illumination . Figure 10 shows the response to an electric shock of 6 μA applied during 1 s. We observe an increase of about 7% of the fluorescence emission after the electric shock delivery which corresponds to a transient local increase of the Ca2+ concentration in the α branch of the mushroom bodies (see  for drosophila brain structure details). This good signal-to-noise ratio was performed even though the exposure time for every 512 × 512 pixels uniform and structured images was only 10 ms. The frame rate of 30 Hz is similar to video rate. We are able to decrease technologically again this exposure time because our microscope is based on the wide-field technique and the DLP can perform in the microsecond range the switch between uniform and structured illuminations. The main limitation for our experiment remains the signal-to-noise ratio. Typically confocal microscopy imaging in drosophila brain are in the Hz range. Five hertz frame rate confocal imaging were used for example to focus on the mnesic traces in α and β MB neurons after conditioning  or to study the ability of Kenyon cells dendrites to release synaptic vesicles  with images of relatively small resolution.
We implemented a new fast optical microscope based on the structured illumination by a digital micromirror device. After optimization and suppression of coherent artefacts observed with a laser light source, we obtained a diffraction limited spatial resolution laterally and axially. Images recorded on synthetic model samples as well as in vivo brain imaging showed that the developed instrument spatial resolution is close to that of the confocal microscope. It is equivalent laterally and the optical sectioning thickness is slightly larger but close to confocal performance. Nevertheless, our microscope has two major advantages with respect to confocal microscopy. The combination of wide field illumination with the rapidity of micromirror device allowed to acquire images at a rate larger than 30 Hz, that is more than one order of magnitude faster than a typical confocal microscope on the same type of sample for the same spatial resolution (512×512 pixels). As the illumination can be changed rapidly, an advantage of the instrument that we did not use in this study, is to adjust in real time the spatial frequency with the imaging depth in the biological sample. Indeed, as in confocal microscopy the pinhole can be opened to keep the best compromise between resolution and brightness (but not in real time during a stack recording), the spatial frequency can be reduced for the same purpose in structured illumination microscopy because diffusion and aberrations will attenuate the transmission of higher spatial frequencies. In conclusion, the speed and the resolution of the instrument will open the opportunity to analyze more globally the activity of neural networks involved in different biological functions. The targeted biological sample was drosophila brain but such an approach should benefit to other organisms with genetically targeted sparse labeling as zebrafish for example.
This work was supported for Aimé Cotton team by the region Ile-de-France (Sesame contract) and the Federation Lumat.
References and links
1. A. H. Brand and N. Perrimon, “Targeted gene expression as a means of altering cell fates and generating dominant phenotypes,” Development 118, 401–415 (1993). [PubMed]
2. B. D. Pfeiffer, T. T. Ngo, K. L. Hibbard, C. Murphy, A. Jenett, J. W. Truman, and G. M. Rubin, “Refinement of tools for targeted gene expression in Drosophila,” Genetics 186(2), 735–755 (2010). [CrossRef] [PubMed]
3. K. J. Venken and H. J. Bellen, “Transgenesis upgrades for Drosophila melanogaster,” Genetics 134(20), 3571–3584 (2007).
4. G. Miesenbock, D. A. De Angelis, and J. E. Rothman, “Visualizing secretion and synaptic transmission with pH-sensitive green fluorescent proteins,” Nature 394(6689), 192–195 (1998). [CrossRef] [PubMed]
5. A. Fiala, T. Spall, S. Diegelmann, B. Eisermann, S. Sachse, J. M. Devaud, E. Buchner, and C. G. Galizia, “Genetically expressed cameleon in Drosophila melanogaster is used to visualize olfactory information in projection neurons,” Current Biology 12(21), 1877–1884 (2002). [CrossRef] [PubMed]
6. Y. Wang, H. F. Guo, T. A. Pologruto, F. Hannan, I. Hakker, K. Svoboda, and Y. Zhong, “Stereotyped odor-evoked activity in the mushroom body of Drosophila revealed by green fluorescent protein-based Ca2+ imaging,” J Neurosci. 24(29), 6507–6514 (2004). [CrossRef] [PubMed]
8. N. Gervasi, P. Tchenio, and T. Preat, “PKA dynamics in a Drosophila learning center: coincidence detection by rutabaga adenylyl cyclase and spatial regulation by dunce phosphodiesterase,” Neuron 65(4), 516–529 (2010). [CrossRef] [PubMed]
10. R. H. Webb, “Confocal optical microscopy,” Rep. Prog. Phys. 59, 427–471 (1996). [CrossRef]
12. A. Bullen, S. S. Patel, and P. Saggau, “High-speed, random-access fluorescence microscopy: High-resolution optical recording with voltage-sensitive dyes and ion indicators,” Biophys J. 73(1), 477–491 (1997). [CrossRef] [PubMed]
13. G. Q. Xiao and G.S. Kino, “A real-time confocal scanning optical microscope,” Proc. SPIE 0809, 107–113(1987). [CrossRef]
14. J. Bewersdorf, R. Pick, and S. W. Hell, “Multifocal multiphoton microscopy,” Optics Letters 23(9), 655–657 (1998). [CrossRef]
16. V. Nikolenko, B. O. Watson, R. Araya, A. Woodruff, D. S. Peterka, and R. Yuste, “SLM microscopy: scanless two-photon imaging and photostimulation with spatial light modulators,” Front. Neural Circuits 2(5), 1–14 (2008). [CrossRef]
18. J. Huisken, J. Swoger, F. Del Bene, J. Wittbrodt, and E. H. K. Stelzer, “Optical sectioning deep inside live embryos by selective plane illumination microscopy,” Science 305(5686), 1007–1009 (2004). [CrossRef] [PubMed]
19. M. A. A. Neil, R. Juskaitis, and T. Wilson, “Method of obtaining optical sectioning by using structured light in a conventional microscope,” Opt. Express 22(24) 1905–1907 (1997).
20. L. H. Schaeffer, D. Schuster, and J. Schaffer, “Structured illumination microscopy: artefact analysis and reduction utilizing a parameter optimization approach,” J Microsc. 216(2), 165–174 (2004). [CrossRef]
21. D. Lim, K. K. Chu, and J. Mertz, “Wide-field fluorescence sectioning with hybrid speckle and uniform-illumination microscopy,” Opt. Express 33(16), 1819–1821 (2008).
23. S. Santos, K. K. Chu, D. Lim, N. Bozinovic, T. N. Ford, C. Hourtoule, A. C. Bartoo, S. K. Singh, and J. Mertz, “Optically Sectioned Fluorescence Endomicroscopy with Hybrid-Illumination Imaging through a Flexible Fiber Bundle,” Advances in Imaging 14(3), 30502 (2009).
24. T. N. Ford, D. Lim, and J. Mertz, “Fast optically sectioned fluorescence HiLo endomicroscopy,” J Biomed Opt. 13(2), 021105 (2012). [CrossRef]
25. B. Zang, J. Zerubia, and J. C. Olivo-Marin, “Gaussian approximations of fluorescence microscope point-spread function models,” Applied Optics 46(10), 1819–1829 (2007). [CrossRef]
26. J. B. Pawley, Handbook of Biological Confocal Microscopy (Plenum, 2006). [CrossRef]
27. J. W. Goodman, Introduction to Fourier Optics, 3rd ed. (Roberts, 2005).
28. A. Stockseth, “Properties of a Defocused Optical System,” JOSA 59(10), 1314–1321 (1969). [CrossRef]
30. M. Heisenberg, A. Borst, S. Wagner, and D. Byers, “Drosophila mushroom body mutants are deficient in olfactory learning,” J Neurogenet. 13, 1–30 (1985). [CrossRef]
31. J. B. Duffy, “GAL4 system in Drosophila: a fly geneticist’s Swiss army knife,” Genesis 34(1–2), 516–529 (2002). [CrossRef]
32. R. I. Wilson, G. C. Turner, and G. Laurent, “Transformation of Olfactory Representations in the Drosophila Antennal Lobe,” Science 303(5656), 366–370 (2004). [CrossRef]
33. L. Tian, S. A. Hires, T. Mao, D. Huber, M. E. Chiappe, S. H. Chalasani, L. Petreanu, J. Akerboom, S. A. Kinney, E. R. Schreiter, C. I. Bargmann, V. Jayaraman, K. Svoboda, and L. L. Looger, “Imaging neural activity in worms, flies and mice with improved GCaMP calcium indicators,” Nature Methods 6(12), 875–881 (2009). [CrossRef] [PubMed]
34. D. Yu, D. B. G. Akalal, and R. L. Davis, “Drosophila alpha/beta mushroom body neurons form a branch-specific, long-term cellular memory trace after spaced olfactory conditioning,” Neuron 52(5), 845–855 (2006). [CrossRef] [PubMed]
35. F. Christiansen, C. Zube, T. F. M. Andlauer, C. Wichmann, W. Fouquet, D. Owald, S. Mertel, F. Leiss, G. Tavosanis, A. J. F. Luna, A. Fiala, and S. Sigrist, “Presynapses in Kenyon Cell Dendrites in the Mushroom Body Calyx of Drosophila,” J. Neurosci. 31(26), 9696–9707 (2011). [CrossRef] [PubMed]