An experimental X-ray cone-beam microtomographic imaging system utilizing a generalized Feldkamp reconstruction algorithm has been developed in our laboratory. This microtomographic imaging system consists of a conventional dental X-ray source (Aztech 65, Boulder, CO), a sample position and rotation stage, an X-ray scintillation phosphor screen, and a high resolution slow scan cooled CCD camera (Kodak KAF 1400). A generalized Feldkamp cone-beam algorithm was used to perform tomographic reconstruction from cone-beam projection data. This algorithm was developed for various hardware configuration to perform reconstruction of spherical, rod-shaped and plate-like specimen.

A test sample consists of 8 glass beads (approx. 800μm in diameter) dispersed in an epoxy-filled #0 gelatin capsule. One hundred X-ray projection images were captured equal angularly (at 3.6 degree spacing) by the cooled CCD camera at a of 1317×967 (17×17mm2) pixels with 12-bit dynamic range. Figure 1 shows a 3D isosurface rendering of the test sample. The eight glass beads and trapped air bubbles (arrows) in the epoxy resin (e) are clearly visible.

We are developing a collaborative research environment, the Collaboratory for Microscopic Digital Anatomy (CMDA), to provide remote access to the sophisticated instrumentation located at the Na-tional Center for Microscopy and Imaging Research (NCMIR). The project’s initial focus is the col-lection and analysis of data from NCMIR’s unique intermediate-high voltage transmission electron microscope (HVEM), an instrument expressly designed to obtain images from thick specimens con-taining substantial 3-D structure. Because of the electron optical characteristics of the microscope, its images represent a 2-D projection of the specimen’s 3-D structure. 3-D data is derived, using axial tomography, from a series of projections acquired as the specimen is successively tilted in small angu-lar increments. Visualizing the 3-D volume data generated by this procedure is a key challenge facing the project. Our experience suggests that existing visualization mechanisms are limited in their ability to fully access the data’s biologically interesting information.

The compound eye of the fruit fly, Drosophila melanogaster, is composed of a highly ordered array of facets (FIG. 1), each containing a precise set of neurons and supporting cells. The eye arises during the third larval instar from an undifferentiated epithelium, the eye imaginai disc, which is connected to the brain via the optic stalk (FIG. 2). During eye development, movement of the morphogenetic furrow, progressive recruitment of specific cell types and the growth of photoreceptor axons into the brain are each dynamic processes that are routinely studied indirectly in fixed tissues. While stereotyped development and the ‘crystalline’ like structure of the eye facilitates this analysis, certain experiments are hindered by the inability to observe developmental processes as they occur. To overcome this limitation, we have combined organ culture with advanced microscopy tools to enable the observation of eye development in living tissue.

We are using Data Explorer (DX), a general-purpose, interactive visualization program developed by IBM, to perform three-dimensional reconstructions of neural structures from microscopic or optical sections. We use the program on a Silicon Graphics workstation; it also can run on Sun, IBM RS/6000, and Hewlett Packard workstations. DX comprises modular building blocks that the user assembles into data-flow networks for specific uses. Many modules come with the program, but others, written by users (including ourselves), are continually being added and are available at the DX ftp site, http://www.tc.cornell.edu/DXhttp://www.nice.org.uk/page.aspx?o=43210.

Initally, our efforts were aimed at developing methods for isosurface- and volume-rendering of structures visible in three-dimensional stacks of optical sections of insect brains gathered on our Bio-Rad MRC-600 laser scanning confocal microscope. We also wanted to be able to merge two 3-D data sets (collected on two different photomultiplier channels) and to display them at various angles of view.

This presentation introduces the current work of the Bioimage Corsortium and its international collaborators in the setting up of a new data base containing volumes of biological specimens.

The scope of the data that we will be organizing during the next three years is very broad, ranging from macromolecules to larger structures or even entire cells. The approach is designed to combine the desire to break many undesired formal barriers between the type of volume information produced at the different resolution levels whith, at the same time, the clear need of implementational flexibility required to appropriately handle different types of data.

The internal organization of the Bioimage Consortium clearly reflects the objectives presented above. In the first place, partners in the consortium have the scientific knowledge and projection in areas ranging from macromolecular structure to cell dynamics. These partners are then nucleated around two well-defined “servers”, one in Madrid, responsible for the work on macromolecular structures, and the other in Heildeberg, responsible for data more in the area of cell biology.

Approaches for exploration of biological structures with modern 3D microscopy techniques are expanding rapidly. These technological advances include not only the imaging of structures, but also the computational reconstruction methods for calculating three-dimensional structures. The computational techniques used to investigate the three-dimensional structure of macromolecular complexes, organelles or cells include crystallographic reconstruction, helical reconstruction, icosahedral reconstruction, single particle reconstruction, electron tomography and serial section reconstruction. These techniques have particularly useful in determining structures which are too small for light microscopy or too large for NMR or X-Ray structure determination or structures for which crystals are not available or order in the crystals is limited.

Advances in microscopes and associated computational capabilities have been substantial over the last few years. In this session, presentations will examine technical aspects and methods for three-dimensional reconstruction of biological structures from images acquired by electron microscopy and analyzed using the computational reconstruction techniques listed above.

The correct reconstruction of the 3D structure from electron micrographs of sets of individual macromolecules or macromolecular assemblies depends crucially on the alignment and orientation determination of the individual images. The lateral alignment from one image to the next is generally easily accomplished by calculating the centre of mass of the individual molecular micrographs. For relatively noise-free pairs of images of unknown structures the sinogram correlation function provides common axes which can determine a correct angular orientation for any set of three images, building consecutively to a complete orientation determination for all images [1]. Low-dose images, however, are inherently noisy. For noisy images the angle determination becomes more problematical, since any fourth noisy image will have three different orientations, each with respect to one of the three possible pairs of three previously oriented noisy images. As more noisy images are added to the set, the number of possible orientations for any given image grow factorially.

The ultimate goal of high resolution electron microscopy is to determine quantitatively the atomic structure of an object. In this respect the electron microscope can be considered as an information channel that carries this information from the object to the observer. High resolution images are then to be considered as data planes from which the structural information has to be extracted.

However this structural information is usually hidden in the images and cannot easily be assessed. Therefore, a quantitative approach is required in which all steps in the imaging process are taken into account. Two main approaches have been followed so far in the literature: the indirect approach in which the images are simulated for various plausible trial structures of the object and compared with the experimental images, and the direct approach in which the lost phase information is retrieved using holographic techniques so as to “deblur” the effect of the microscope and to reveal directly the atomic structure of the object.

Single crystal X-ray diffraction is the traditional method for accurate crystal structure determination. A major difficulty in X-ray crystallography is the phase problem; diffracted intensities contain amplitude information but no phases. In order to solve a structure, the phases of at least the strongest reflections must be estimated by Patterson techniques, so-called direct methods or in any other way. Once the structure has been solved (i.e. the atoms found to within about 0.2 Ångström of their correct positions), then refinement is rather straight-forward for single crystal data. Typically, single crystals diffract to about 1 Â resolution. After refinement, the atomic co-ordinates are obtained with an accuracy of about 0.01 Â for organic molecules and down to 0.001 Â for inorganic structures. One limitation of single crystal X-ray diffraction is that the crystals need to be at least about 10μm in all dimensions in order to diffract, even if the radiation source is a synchrotron.

Several methods have been suggested for elliminating the twin image in in-line holography, including Franhofer holography as used in our previous work, integration over focus, integration over wavelength, integration over lateral source position and recursion methods such as Gerchberg-Saxton or Fienup (see for all refs.). Fraunhofer holography requires the Fresnel number N =, with d the object size. The integration methods depend on the fact that the real and virtual images have different dependencies on some parameter such as focus, and have proved powerful in photoemission holographic diffraction. Here we apply focus integration and recursion to low voltage, point-projection images of carbon films, and discuss the effects of inhomogeneities in the electric field which arise when the tip is very close to a fibre-shaped sample such as DNA. by embedding biomolecules in solid Xenon at 55K, and using a beam energy of about 9 volts (less than the Xenon bandgap minus the electron affinity), radiation damage to the molecule may be minimized, while the molecule is gently supported by an almost electron transparent medium.

Fractal dimension is useful for characterizing objects with tortuous boundaries, providing an additional measurement to characterize shape beyond the commonly used area, perimeter, caliper diameter and their combinations. The fractal dimension of the boundary describes the increase of the measured value of the perimeter of an object resulting from increasing measurement resolution, i.e., decreasing step size. We desire to characterize and classify shapes such as soot particles, as shown in the TEM micrograph in Fig. 1a, and crystalline grains of metal alloys, as shown in the SEM micrograph of Raney nickel in Fig. 4.

This investigation involves evaluation of the Minkowski-sausage or dilation technique (Fig. 1) using the histogram of the Euclidean distance map of the perimeter of the object. The distance map gives bands of various widths that follow the perimeter - greater widths lose detail. Perimeter length is sausage (band) area divided by sausage width, w.

Z-contrast produces an image that varies strongly with atomic number and offers limited chemical sensitivity. The annular dark field scanning transmission microscope (ADF-STEM) has recently been used to produce Z-contrast images at high resolution. In a STEM the electron beam is focused into a small probe at the specimen plane and scanned across the specimen. The electrons scattered at high angle are collected by an annular detector to produce a signal that approximately varies as Z11.5 to Z2. The only commercial supplier of dedicated STEM'S (VG-Microscopes, now Thermo) has recently discontinued production, so it is worth considering how to produce an equivalent Z-contrast image in a conventional transmission electron microscope (CTEM).

The reciprocity theorem states that the STEM and CTEM are equivalent if the source and detector are interchanged (reciprocity hold for all orders of elastic scattering so it is not necessary to restrict this discussion to weakly scattering specimens).

Annular dark-field (ADF) imaging in the scanning transmission electron microscope (STEM) at atomic resolution can be regarded as being almost perfect incoherent imaging, which has two major advantages over conventional high-resolution transmission electron microscopy (HRTEM), which is close to being perfectly coherent: Firstly, the images formed are direct structure images of the projected atomic structure, with regions of intensity located at the positions of the atomic columns. Secondly, the images can be written as the convolution between two real and positive functions: a point-spread function given by the intensity of the STEM probe, and an object function that consists of narrow, almost <5-function-like, peaks located exactly over the nuclei of the atomic columns. Thus there is no phase problem in ADF STEM imaging, unlike in conventional HRTEM where phase retrieval techniques such as holography are often employed.

The lack of a phase problem in ADF STEM imaging creates immediate opportunities for image processing.

The advance of digital recording and energy filtering now makes it possible to extract quantitative crystal structure information in an accuracy close to or better than the x-ray method. The extraction of structure information employs the refinement technique by comparing theory and experiment using chisquare criteria of

For the unit cell or crystal orientation determination, the experiment yx is the line equation of Kikuchi or HOLZ lines, and/or positions of convergent beam disks or diffraction spots. Lines, or disks, can be measured directly from the digitally recorded two-dimensional patterns using techniques of image processing and transformation such as Hough. A practical method for measuring unit cell parameters was described by [1]. For the determination of atomic positions, the experiment yx is the diffracted intensities of medium and high order reflections. For the measurement of charge density, the yx is the energy-filtered diffracted intensities of low order reflections or high order reflections coupled with a low order reflections.

Convergent beam electron diffraction (CBED) produces a diffraction pattern from a single point on the specimen using a focused probe in the scanning transmission electron microscope (STEM). Because the incident wavefunction is a focused probe, each diffraction spot is enlarged to a disk the size of the objective aperture. Thermal vibration in the specimen reduces the intensity of the diffraction disks and introduces scattering in between the disk. This normally forbidden scattering between the diffraction disks is referred to as thermal diffuse scattering (TDS). The effects are most pronounced at large scattering angles, where TDS scattering can greatly reduce the intensity of the higher order Laue zone (HOLZ) lines. Previous work on modeling the TDS intensity involved the so-called frozen phonon method and the Einstein model of lattice vibration. The Einstein model assumes that each atom in the specimen vibrates independently of every other atom in the specimen and possible correlations among adjacent atoms are ignored.

In recording the intensity of the diffraction pattern from an arbitrary object, the phase is lost. This gives rise to the “phase problem”in reconstructing the object in crystallography, where the pattern is sampled at the Bragg peaks. It was shown by Bates and by Hayes that sampling at twice crystallographic density (2x oversampling) in each dimension is sufficient in principle to obtain a full reconstruction of the object. It was pointed out by Sayre in 1991 that the diffraction pattern of a non-crystalline object is continuous, therefore the pattern may be sampled on a finer scale, providing the additional information for the image reconstruction.

The loss of phase corresponds to a loss of half the information. It seems therefore that 2x oversampling in each dimension (8x in three dimension) may be unnecessary. We find that oversampling 1.67x in each dimension is sufficient to obtain a reconstruction, even in the presence of noise. However our algorithm fails when we reduce the oversampling factor to 1.58x.

The design of modern magnetic storage devices with their rapidly increasing information density ne-cessitates a full characterization of the underlying microstructure of the device materials. Lorentz microscopy has for several decades been the main vehicle for micro-magnetic observations. The pri-mary limitation of the Lorentz modes is perhaps the attainable magnification. To avoid saturation, the sample must be placed in a low-field or preferably field-free region in the column, and this invariably means that the microscope can only be operated at low magnifications. Both Fresnel and Foucault images are rather sensitive to the exact experimental conditions which renders quantitative observa-tions quite difficult, if not impossible. Inelastic scattering further limits the usefulness of Lorentz observations to very thin foils.

We have recently reported1 a novel Lorentz microscopy setup, combining a Gatan Imaging Filter (GIF) and a JEOL 4000EX top-entry high resolution TEM, operated at 400 kV with the main objective lens switched off.

1.Introduction:If digital image matching between an experimental HREM-image and the simulated image fails, one of the suggested reasons is the “inaccuracy of the atomic form factors“ describing the electron scattering in common simulation packages using the free and neutral atom approximation from Hartree-Fock calculations [1]. In detail, three contributions within the usual form factor calculations are mainly missing: (i) the redistribution of charge in ionic crystals, (ii) the accumulation of charge away from atom sites in covalent crystals, (iii) the inclusion of thermal diffuse scattering (TDS) as well as the correlated vibration of atoms beyond the Einstein-approximation within the Debye-Waller factor (DWF) theory. Each of the three effects have been checked separately:

2.Simulations of TDS by Coupling of Molecular Dynamics Time Series to HREM- Multislice Calculations:70 snap shots of a series of structures of NiAl (4.9 × 5.1 × l0nm) are stored from an equilibrated molecular dynamics simulation with vibration amplitudes corresponding to room temperature.

Teaching of transmission electron microscopy usually includes both lectures on the contrast theories, electron diffraction, etc., and practical hands-on operation of the microscope. The number of students attending the lectures is normally unlimited, but at the microscope, only a few persons can work at the same time. Since the microscopes are expensive, it would be of a great help if cheaper 'training' microscopes with basic imaging and diffraction capabilities were available. These functions, in fact, can quite easily be realized with fast personal computers and work stations, where the simulation of transmission electron micrographs and related diffraction patterns can help the student better understand the image formation processes. Adding text, audio and video help capabilities to the program, it can be made an efficient supplemental teaching tool.

TemTutor for Windows is based on microScope for Windows, which is a BF/DF TEM micrograph simulation program for dislocations and stacking faults.

The most useful classical definition of resolution is a property of an instrument alone, independent of the sample. Because of this it is not possible to define resolution simply for HREM or for most of the new microscopies. For weak phase objects, the best current HREM machines offer a point resolution very close to 0.lnm . For the first time, recent off-axis electron holography has extended resolution beyond the point resolution limit, in fulfillment of Gabor’s original aim. Other recent approaches to super-resolution have used through-focus series under highly coherent conditions, sums of images obtained under different illumination angles, dark-field STEM imaging with high-angle detectors,thin annular detectors in STEM, or the use of coherent nanodiffraction patterns. Aberration correction by electron optical means has undergone a recent revival, and the effects of third-order astigmatism are under study. If third-order spherical aberration is corrected, the projected charge density approximation may be more useful than the weak-phase object approximation, since the Scherzer focus then no longer applies.