Microsc. Microanal. 21, 172–178, 2015 doi:10.1017/S1431927614013610

© MICROSCOPY SOCIETY OF AMERICA 2014

Imaging Samples in Silica Aerogel Using an Experimental Point Spread Function Amanda J. White * and Denton S. Ebel Department of Earth and Planetary Sciences, American Museum of Natural History, Central Park West at 79th Street, New York, NY 10024, USA

Abstract: Light microscopy is a powerful tool that allows for many types of samples to be examined in a rapid, easy, and nondestructive manner. Subsequent image analysis, however, is compromised by distortion of signal by instrument optics. Deconvolution of images prior to analysis allows for the recovery of lost information by procedures that utilize either a theoretically or experimentally calculated point spread function (PSF). Using a laser scanning confocal microscope (LSCM), we have imaged whole impact tracks of comet particles captured in silica aerogel, a low density, porous SiO2 solid, by the NASA Stardust mission. In order to understand the dynamical interactions between the particles and the aerogel, precise grain location and track volume measurement are required. We report a method for measuring an experimental PSF suitable for threedimensional deconvolution of imaged particles in aerogel. Using fluorescent beads manufactured into Stardust flight-grade aerogel, we have applied a deconvolution technique standard in the biological sciences to confocal images of whole Stardust tracks. The incorporation of an experimentally measured PSF allows for better quantitative measurements of the size and location of single grains in aerogel and more accurate measurements of track morphology. Key words: point spread function, deconvolution, aerogel, Stardust, confocal

I NTRODUCTION The NASA Stardust mission returned to Earth with material from comet 81 P/Wild 2 in 2006. During the mission, incoming cometary particles were captured by impact into a silica aerogel collector tray at a relative velocity of 6.1 km/s (Brownlee et al., 2006). Each collection event represents a unique hypervelocity impact resulting in a three-dimensional (3D) track. The nature of each track-forming event, including the original state of the impactor, is recorded in the 3D track morphology. By characterizing the unique 3D structure of whole Stardust tracks, we can begin to infer properties such as the size, shape, and porosity of the original impactors. Confocal microscopy is a powerful imaging technique that allows for rapid acquisition of 3D data sets and has previously been shown to be well-suited for nondestructively imaging whole Stardust tracks (Kearsley et al., 2007; Ebel et al., 2009; Greenberg & Ebel, 2010, 2012). However, due to limitations inherent to microscopy, the instrument optics convolve with the incoming signal, obscuring information and hindering the ability to locate single grain fragments and identify true track and particle size and shape. This information is necessary for understanding the original impactor properties. For this reason, it is essential to deconvolve confocal data in order to recover lost information and to make meaningful quantitative measurements. In this paper, we will discuss two methods of deconvolution Received June 25, 2014; accepted October 28, 2014 *Corresponding author. [email protected]

and their application to samples returned by the NASA Stardust mission. A central goal of our work is to study the morphological structure and distribution of both cometary material and silica aerogel (compressed and uncompressed) along whole Stardust tracks. Silica aerogel is a porous, low density, SiO2 solid. Because of its ultra-low density and fine mesostructure, aerogel is an effective capture medium for hypervelocity particles (Tsou, 1995), which has been used successfully in particle capture missions. Aerogel is a potential material for future missions (Tsou et al., 2012; Jones et al., 2014); therefore it is crucial to develop the ability to correctly identify particles in aerogel. As aerogel is translucent, optical methods can be utilized and morphological studies can be performed using laser scanning confocal microscopy (LSCM). A confocal microscope functions by using a pinhole aperture to block all unfocused light making it possible to nondestructively image a sample by raster scanning slices at successive focal depths and to quickly obtain a 3D image stack of an entire particle track (Minsky, 1961; Inoué, 2006). Confocal microscopy as it relates to Stardust samples in particular is further described by Greenberg & Ebel (2010). Deconvolution is an image processing step that reduces or reverses distortion introduced by the imaging system (Parton & Davis, 2006). Deconvolution is required to remove distortion along the Z axis (optic axis) of 3D images before reliable quantitative measurements of fine, detailed structures can be made. This systematic Z-axis aberration is inherent to the configuration of the optical path in any microscope, including the path through the experimental

Imaging Samples in Silica Aerogel

setup (e.g., aerogel). Deconvolution methods attempt to restore convoluted images using a point spread function (PSF), a mathematical image representation of the optical distortion of a single point by the microscope optics. A PSF can be specified using multiple techniques including (1) a theoretical calculation based on the instrument’s physical and optical properties or (2) a series of experimental measurements. A theoretical PSF is a good approximation for correcting optical distortion, however, it only takes into account the best alignment and imaging conditions on a given microscope and does not account for any deviations from an ideal optical path. A measured PSF can be determined from images of very small particles, such as latex beads, of a known shape and size that are smaller than the laser wavelength used to image them (Cannell et al., 2006). Because the bead images used to derive the measured PSF are obtained under the same conditions as images of the sample of interest, the measured PSF can accurately account for more variables of the experimental configuration than the theoretical PSF. This results in a dramatic difference in the shape and size of the measured and theoretical PSFs as seen in Figure 1. A measured PSF is typically >20% larger and asymmetric when compared with a theoretical PSF for the same data set

173

(Cannell et al., 2006). In our previous work on Stardust tracks (Greenberg & Ebel, 2010, 2012), we have used a theoretical PSF. The measured experimental PSF (Fig. 1b) is highly elongated in the direction of the optical axis. This implies that the alignment of lenses in our instrument may deviate from the optimal setup (Cannell et al., 2006). The elongation is also, at least partially, because of a mismatch between the refractive indices of air and aerogel. The difference of refractive indices causes distortions in the PSF and decreases the resolution along the Z axis; however, it is not accounted for in the theoretical calculation. In the case of particles captured in aerogel, the aerogel itself constitutes a deviation from ideality as the PSF calculation assumes samples are in air. Using a set of custom made keystones, thin triangle-shaped prisms (Westphal et al., 2004), of flightgrade aerogel with 100 nm fluorescent beads mixed into the precursor solution, we have measured a PSF for use in imaging Stardust tracks in aerogel.

MATERIALS

AND

METHODS

The optimal material for obtaining an experimental aerogel PSF would be manufactured in the same batch as the aerogel

Figure 1. Side-by-side, same scale comparison of a theoretical (a) and a measured (b) point spread function (PSF) in the XZ plane. The measured PSF is noticeably larger and more elongated along the optical axis (Z-direction) of the instrument. The measured PSF is also less symmetrical than its theoretical counterpart. These differences stem from the fact that more of the experimental variables are accounted for in the measured PSF. Here and in all other PSFs the color scheme is an arbitrary lookup table applied to the image by the Huygens software for clarity. Intensity values vary from 0 to 10 − 2.

174

Amanda J. White and Denton S. Ebel

Figure 2. A side-by-side, same scale comparison of a 500 nm fluorescent bead image before it has been processed (a), and after being deconvolved using a measured point spread function (PSF) (b) and a theoretical PSF (c). Both deconvolutions improve image contrast. The measured PSF (b) removes more structure from the image, which is thought to be an artifact caused by reflections off background aerogel. Intensities peak at 6 × 104 in (a), 5 × 105 in (b), and 105 in (c).

to which the PSF will be applied. In order to create an applicable experimental PSF for use with Stardust tracks, custom aerogels were made with fluorescent TetraSpeck latex beads, manufactured by Life Technologies (Carlsbad, CA, USA), mixed into the precursor solution. These aerogels

were manufactured by S. M. Jones (NASA Jet Propulsion Laboratory) in 2011 using the same method as was used (c. 1998) for flight-grade aerogel for the Stardust mission. Two such batches of aerogel were made, one containing 100 nm beads (T7279) and one containing 500 nm beads (T7281).

Imaging Samples in Silica Aerogel

The aerogels were then cut into keystones and mounted on glass microforks by D. R. Frank (NASA Johnson Space Center) in the same fashion as a Stardust sample track.

175

Using an identical scan configuration as that used for Stardust tracks allocated to the authors, bead-bearing keystones were imaged at the Microscopy and Imaging Facility

Figure 3. Intensity profiles of the deconvolved and raw data in the Z plane (left) and X plane (right). Y plane data is nearly identical to the X plane. Both deconvolution techniques resulted in a sharper peak and smaller full-width at half-maximum relative to the raw data. This indicates the measured point spread function (PSF) restored much more intensity and positional accuracy to the image than did the theoretical PSF.

Figure 4. Measured point spread function for the stand-alone case of 500 nm latex beads on glass in both the XY plane (a) and the XZ plane (b). Intensity peaks at 3 × 10 − 4.

176

Amanda J. White and Denton S. Ebel

Figure 5. A sample image displaying two-dimensional (2D) projections of 3D confocal maps of Track 131 (C2012,7,131,0,0) in the original data (a) and after deconvolution using an experimental PSF (b). Particle locations are much more definitive in the second map making it more useful for volume and grain analysis.

Imaging Samples in Silica Aerogel

of the American Museum of Natural History using a Zeiss LSM 710 LSCM (Göttingen, Germany). High-resolution 3D, reflected light image stacks were acquired using a 488 nm laser at a voxel size of 74 × 74 nm in X–Y and 360 nm in Z (optical axis). This resolution was selected in order to obtain ideal sampling rates as outlined by the Nyquist criterion (Pawley, 2006a, 2006b). Images from the keystone containing 100 nm beads were combined and distilled into a measured PSF using SVI’s Huygens Professional v4.2 Software (SVI, 2014). For 3D deconvolution of raw image data, we use SVI’s Huygens Professional v4.2 software that employs an iterative, classic maximum likelihood estimation (CMLE) method to deconvolve images. CMLE methods attempt to reassign out of focus light and recover some lost information (Parton & Davis, 2006; SVI, 2014). Following a reviewer’s suggestion, we imaged dry 500 nm latex beads transferred to a glass slide in distilled water. These stand-alone bead images were acquired using 405, 561 and 633 nm lasers, at 74 × 74 × 360 nm (X–Y–Z) resolution. In this case, the images represent bead fluorescence rather than reflected light. The beads do not fluoresce at the 488 nm wavelength, which we found to be optimal for imaging particles by reflection in aerogel.

RESULTS AND D ISCUSSION In order to compare the effectiveness of both the measured and theoretical PSFs, we imaged an aerogel keystone containing 500 nm beads using the same LSCM scan parameters that were used to acquire the measured PSF derived from 100 nm beads in aerogel. Deconvolutions using both measured and theoretical PSFs were then performed on this data. The image used in these tests consisted of an average of 17 different beads (ten images of individual beads and one large image of seven beads), which were imaged at different depths and locations within the keystone. This was done in order to reduce the effects of the background aerogel in images, limit photon noise, and to better account for light/signal drop-off at deeper locations in the keystone. Deconvolution is expected to increase image contrast and to bring out small spatial features by removing larger, out of focus features. This can be seen in Figure 2, which shows the 3D center of the initial raw data (a) alongside the results of both deconvolution tests (b and c). Comparing the raw data (Fig. 2a) with the results of a deconvolution using a theoretically determined PSF (Fig. 2c), one can observe how the deconvolution enhances some features from the original image; however, these features are not necessarily real. The structure is most likely an effect of light being refracted off the background aerogel, a deviation from ideality in our optical setup, or a combination of the two. The image of a 500 nm bead is qualitatively better represented by the results of the deconvolution performed using a measured PSF (Fig. 2b). Both the optical setup of the LSM 710 and the effects of the aerogel are better accounted for by the measured PSF than by the theoretical PSF. This is further shown in Figure 3 that displays the intensity profiles along

177

the Z and X plane for the raw data as well as the same data processed using both theoretical and experimental deconvolution methods. In both deconvolution tests, the intensity peaks were sharper than in the raw data and a decrease in the full-width at half-maximum (FWHM) was observed. The measured PSF restored much more of the image’s intensity than the theoretical PSF did. As the FWHM is a good estimation of spatial resolution (SVI, 2014), the sharper peak also indicates a greater positional accuracy for resolved grains. The measured PSF for the images of bare 500 nm latex beads (Fig. 4) differs slightly from the measured PSF of the 100 nm beads embedded in aerogel. This difference can be attributed to the aerogel itself. However, the direct application of this measurement to aerogel in other contexts is fraught with difficulty. Even in the same batch, aerogels can differ in density, opacity, refractive index, and other properties at centimeter scales (Buzykaev et al., 1999). This measured PSF is unlikely to be directly applicable to other aerogels in other optical instruments.

CONCLUSION While determination of a measured PSF for correction of 3D comet Wild 2 track images has been time consuming, it will allow for significant improvements to the quality of final image products. This will, in turn, allow much more accurate measurements of track properties that are essential for inferring original impactor properties by reverse modeling. An example track, Track 131, is shown in Figure 5 as a 2D projection of the 3D map data. In the undeconvolved map (left), individual particles are seen; however, their location is blurred and unfocused. In the deconvolved map (right) individual particles are clearly defined. We have measured, for the first time, a PSF for particles in aerogel. Aerogel has been used successfully in past particle capture missions and will potentially be used in future missions making correct location and identification of particles in aerogel important. This work informs future analysis of such mission samples. Future missions would benefit from the simultaneous manufacture of bead-bearing aerogel for obtaining experimental PSFs in the microscopes used to image returned samples.

ACKNOWLEDGMENTS The authors thank an anonymous reviewer for extraordinarily helpful comments and suggestions. This work was accomplished with support from NASA LARS grant NNX12AF28G (DSE). The LSM 710 confocal microscope was purchased with support from the American Museum of Natural History and NASA LARS grant NNXIOAH06G (DSE). The authors thank Rebecca Rudolph for helping to obtain the LSCM and James Thostenson (AMNH), David R. Frank (JSC), Steven M. Jones (JPL), and Michael E. Zolensky (JSC) for technical assistance. This research has made use of NASA’s Astrophysics Data System Bibliographic Services.

178

Amanda J. White and Denton S. Ebel

REFERENCES BROWNLEE, D., TSOU, P., ALÉON, J., ALEXANDER, C., ARAKI, T., BAJT, S., BARATTA, G., BASTIEN, R., BLAND, P., BLEUET, P., BORG, J., BRADLEY, J., BREARLEY, A., BRENKER, F., BRENNAN, S., BRIDGES, J., BROWNING, N., BRUCATO, J., BULLOCK, E., BURCHELL, M., BUSEMANN, H., BUTTERWORTH, A., CHAUSSIDON, M., CHEUVRONT, A., CHI, M., CINTALA, M., CLARK, B., CLEMETT, S., CODY, G., COLANGELI, L., COOPER, G., CORDIER, P., DAGHLIAN, C., DAI, Z., D’HENDECOURT, L., DJOUADI, Z., DOMINGUEZ, G., DUXBURY, T., DWORKIN, J., EBEL, D., ECONOMOU, T., FAKRA, S., FAIREY, S., FALLON, S., FERRINI, G., FERROIR, T., FLECKENSTEIN, H., FLOSS, C., FLYNN, G., FRANCHI, I., FRIES, M., GAINSFORTH, Z., GALLIEN, J.P., GENGE, M., GILLES, M., GILLET, P., GILMOUR, J., GLAVIN, D., GOUNELLE, M., GRADY, M., GRAHAM, G., GRANT, P., GREEN, S., GROSSEMY, F., GROSSMAN, L., GROSSMAN, J., GUAN, Y., HAGIYA, K., HARVEY, R., HECK, P., HERZOG, G., HOPPE, P., HÖRZ, F., HUTH, J., HUTCHEON, I., IGNATYEV, K., ISHII, H., ITO, M., JACOB, D., JACOBSEN, C., JACOBSEN, S., JONES, S., JOSWIAK, D., JUREWICZ, A., KEARSLEY, A., KELLER, L., KHODJA, H., KILCOYNE, D., KISSEL, J., KROT, A., LANGENHORST, F., LANZIROTTI, A., LE, L., LESHIN, L., LEITNER, J., LEMELLE, L., LEROUX, H., LIU, M.C., LUENING, K., LYON, I., MACPHERSON, G., MARCUS, M., MARHAS, K., MARTY, B., MATRAJT, G., MCKEEGAN, K., MEIBOM, A., MENNELLA, V., MESSENGER, K., MESSENGER, S., MIKOUCHI, T., MOSTEFAOUI, S., NAKAMURA, T., NAKANO, T., NEWVILLE, M., NITTLER, L., OHNISHI, I., OHSUMI, K., OKUDAIRA, K., PAPANASTASSIOU, D., PALMA, R., PALUMBO, M., PEPIN, R., PERKINS, D., PERRONNET, M., PIANETTA, P., RAO, W., RIETMEIJER, F., ROBERT, F., ROST, D., ROTUNDI, A., RYAN, R., SANDFORD, S., SCHWANDT, C., SEE, T., SCHLUTTER, D., SHEFFIELD-PARKER, J., SIMIONOVICI, A., SIMON, S., SITNITSKY, I., SNEAD, C., SPENCER, M., STADERMANN, F., STEELE, A., STEPHAN, T., STROUD, R., SUSINI, J., SUTTON, S., SUZUKI, Y., TAHERI, M., TAYLOR, S., TESLICH, N., TOMEOKA, K., TOMIOKA, N., TOPPANI, A., TRIGO-RODRÍGUEZ, J., TROADEC, D., TSUCHIYAMA, A., TUZZOLINO, A., TYLISZCZAK, T., UESUGI, K., VELBEL, M., VELLENGA, J., VICENZI, E., VINCZE, L., WARREN, J., WEBER, I., WEISBERG, M., WESTPHAL, A., WIRICK, S., WOODEN, D., WOPENKA, B., WOZNIAKIEWICZ, P., WRIGHT, I., YABUTA, H., YANO, H., YOUNG, E., ZARE, N., ZEGA, T., ZIEGLER, K., ZIMMERMAN, L., ZINNER, E. & ZOLENSKY, M. (2006). Comet 81P/Wild 2 under a microscope. Science 314, 1711–1716. BUZYKAEV, A.R., DANILYUK, A.F., GANZHUR, S.F., KRAVCHENKO, E.A. & ONUCHIN, A.P. (1999). Measurement of optical parameters of aerogel. Nucl Instrum Meth Phys Res A 433, 396–400. CANNELL, M.B., MCMORLAND, A. & SOELLER, C. (2006). Image enhancement by deconvolution. In Handbook of Biological Confocal Microscopy, 3rd ed., Pawley, J.B. (Ed.), pp. 488–497. New York: Plenum Press.

EBEL, D.S., GREENBERG, M., RIVERS, M.L. & NEWVILLE, M. (2009). 3-dimensional textural and compositional analysis of particle tracks and fragmentation history in aerogel. Meteor Planet Sci 44, 1445–1463. GREENBERG, M. & EBEL, D.S. (2010). Laser scanning confocal microscopy of Stardust material. Geosphere 6, 515–523. GREENBERG, M. & EBEL, D.S. (2012). Properties of original impactors estimated from three-dimensional analysis of whole Stardust tracks. Meteor Planet Sci 47, 634–648. INOUÉ, S. (2006). Foundations of confocal scanned imaging in light microscopy. In Handbook of Biological Confocal Microscopy, 3rd ed., Pawley, J.B. (Ed.), pp. 1–19. New York: Plenum Press. JONES, S.M., ANDERSON, M.S., DAVIES, A.G., KIRBY, J.P., BURCHELL, M.J. & COLE, M.J. (2014). Aerogel dust capture for in situ mass spectroscopic analysis. Lunar Planet Sci XLV, 2pp. abstract #2104. KEARSLEY, A.T., BALL, A.D., GRAHAM, G.A., BURCHELL, M.J., ISHII, H., COLE, M.J., WOZNIAKIEWICZ, J., HÖRZ, F. & SEE, T.H. (2007). Aerogel track morphology: Measurement, three dimensional reconstruction, and particle location using confocal laser scanning microscopy. Lunar Planet Sci XXXVIII, 2pp. abstract #1690. MINSKY, M. (1961). Microscopy apparatus. US Patent No. 3,013,467. PARTON, R.M. & DAVIS, I. (2006). Lifting the fog: Image restoration by deconvolution. In Cell Biology, 3rd ed., Celis, J.E. (Ed.), pp. 187–200. Burlington: Academic Press. PAWLEY, J.B. (2006a). Fundamental limits in confocal microscopy. In Handbook of Biological Confocal Microscopy, 3rd ed., Pawley, J.B. (Ed.), pp. 20–42. New York: Plenum Press. PAWLEY, J.B. (2006b). Points, pixels, and gray levels: Digitizing image data. In Handbook of Biological Confocal Microscopy, 3rd ed., Pawley, J.B. (Ed.), pp. 59–79. New York: Plenum Press. SVI (2014). SVI-wiki on 3D microscopy, deconvolution, visualization, and analysis. Available at http://www.svi.nl/ FrontPage (accessed November 14, 2014). TSOU, P. (1995). Silica aerogel captures cosmic dust intact. J Non-Cryst Solids 186, 415–427. TSOU, P., BROWNLEE, D.E., MCKAY, C.P., ANBAR, A.D., YANO, H., ALTWEGG, K., BEEGLE, L.W., DISSLY, R., STRANGE, N.J. & KANIK, I. (2012). LIFE: Life investigation for Enceladus a sample return mission concept in search for evidence of life. Astrobiology 12, 730 –742. WESTPHAL, A.J., SNEAD, C., BUTTERWORTH, A., GRAHAM, G.A., BRADLEY, J.P., BAJT, S., GRANT, P.G., BENCH, G., BRENNAN, S. & PIANETTA, P. (2004). Aerogel keystones: Extraction of complete hypervelocity impact events from aerogel collectors. Meteor Planet Sci 39, 1375–1386.

Imaging samples in silica aerogel using an experimental point spread function.

Light microscopy is a powerful tool that allows for many types of samples to be examined in a rapid, easy, and nondestructive manner. Subsequent image...
827KB Sizes 4 Downloads 7 Views