04-09-2012, 01:40 PM
The Best of Both Worlds: 3D X-ray Microscopy with Ultra-high Resolution and a Large Field of View
3D X-ray Microscopy.pdf (Size: 2.09 MB / Downloads: 244)
Abstract.
3D visualizations of complex structures within various samples have been achieved with high spatial
resolution by X-ray computed nanotomography (nano-CT). While high spatial resolution generally comes at the expense
of field of view (FOV). Here we proposed an approach that stitched several 3D volumes together into a single large
volume to significantly increase the size of the FOV while preserving resolution. Combining this with nano-CT, 18-m
FOV with sub-60-nm resolution has been achieved for non-destructive 3D visualization of clustered yeasts that were too
large for a single scan. It shows high promise for imaging other large samples in the future.
INTRODUCTION
X-ray computed tomography (CT) is a powerful tool that non-destructively generates three-dimensional images
of an object’s interior from a series of two-dimensional X-ray images taken around a single axis of rotation. During
the past two decades, the development of advanced x-ray optical devices, such as Fresnel zone plates, coupled with
the high brilliance of synchrotron x-ray sources, has vastly increased the resolution of x-ray imaging. Resolution for
2D imaging down to 12-30 nm has been recently reported [1-4]; meanwhile 3D resolution to 30-60 nm and below is
also now widely available [5-7]. This nano-CT technique has demonstrated many advantages in studying the 3D
structures of complex samples at various length scales and has satisfied applications in a wide range of fields, such
as material science [8-9], cellular biology [10-12], solid oxide fuel cells [13-14], and environmental science [15-17].
However, the high spatial resolution of nano-CT often comes at the expense of field of view (FOV) due, in part,
to limitations in the available number of pixels in a typical CCD-based detector. A typical FOV available for the
highest resolution, such as in the Xradia nanoXCT-S100 at National Synchrotron Radiation Laboratory (NSRL),
Hefei, China, is about 10-15 m [6, 8-9]. These FOV limitations may not be optimal for some studies of larger
samples, in which the volume of interest exceeds the volume available for a single scan.
MATERIALS AND METHODS
Sample Preparation and Nano-CT Experiments
Here, the wild type fission yeast Schizosaccharomyces pombe was used to demonstrate this method. The yeasts
were cultured, fixed, and dehydrated by a process that has previously been proposed for hard X-ray imaging [18].
The samples were then spotted on a silicon nitride membrane of 100-nm thickness for mounting on the microscope
sample holder. The tomography experiments were performed using the Xradia nanoXCT-S100 full-field
transmission hard x-ray microscope (TXM), installed at beamline U7A of NSRL [6]. Operating at 8 keV and
utilizing Zernike phase contrast, this TXM system has been demonstrated to effectively obtain 3D images of some
low-absorbing specimens, such as stained yeasts with about sub-60-nm spatial resolution and 10-m FOV [18]. By
aligning each yeast to the rotation center, two different CT scans were collected with the same exposure dose at
angles ranging from –74º to +74º in 1º intervals using the Xradia TXMController software and reconstructed using
the Xradia TXMReconstructor software [19].
Stitching Algorithm
Due to the experiments being performed using the same sample holder and rotation stage, there are only linear
shifts between the series of tomographic volumes. If the same features are found in each of the two volumes, then
the volumes can be aligned and stitched together into one large volume. A Matlab™ program was developed at
NSRL to provide an automatic stitching process using this method, as shown in Fig. 1. Figures 1(a) and. 1(b) are
two virtual slices of two different 3D volumes, which are each from one large test pattern. One array of dots,
common to both volumes, was located and is indicated by the yellow rectangles. By using linear shifts only, the dots
indicated by the red arrows may be matched to each other and the two volumes aligned. The best linear shifts can be
estimated by a minimum square difference method. The rough coordinations of the same features in two
reconstructed volumes were used as inputs for the Matlab™ program. Then two small cubic volumes with varying
slight liner shifts were cropped and their square differences were calculated. So the minimum difference indicated
the best linear shifts between the two volumes. The two volumes are subsequently stitched together using an
overlapping smoothing function.
CONCLUSION
Here, a stitching method has been proposed to reconstruct volumes larger than the normal FOV of a TXM
system by correlation of different 3D volumes. This technique has been applied to clustered yeast cells that were
thus reconstructed in their entirety with sub-60-nm resolution. The results suggest a novel, useful technique for
studying samples that may normally be limited by the standard FOV, which may extend the potential applications of
commercially available nano-CT instruments in various fields.