features. These approaches are computationally more advanced but variants are
now implemented in some commercial as well as freely available software packages. The reader is referred to the review by Amat et al. [89] for an in-depth
discussion.
After alignment, reconstruction (see following section) is a comparatively simple
task that can be executed via command line, provided that the alignment meets
(user-defined) acceptable standards. For example, the operator might specify a given
tolerance corresponding to sub-pixel alignment. If the pixel size at the detector
corresponds to 1 nm, then sub-pixel alignment corresponds to sub-nanometre
accuracy. This should not be confused with resolution in the final reconstruction. It
simply means that if the alignment accuracy is better than 1 or 2 nm, then it will do
justice to the expected resolution of, say, 4 nm in the final reconstruction. In other
words, it does not degrade resolution.
By definition, the principle of SEM ‘slice and view’ tomography (Chap. 5)
would indicate that individual slices do not require further alignment: the block is
not tilted during acquisition, and the detector-block geometry is maintained
throughout the process, such that sequential images should be in perfect registration
as they are acquired. Specimen drift or other factors such as stage movement can
lead to a translational (xy) shift between one or more slices. These image shifts are
easily corrected during or after acquisition, however, and the process can be
automated. This is done most easily via a cross-correlation function, and it would
make sense to include this in the workflow, even if the unaligned stack passes
visual inspection. Typically, the first step in such a workflow would be to deal with
the massive file size via an initial xy scaling (‘binning’) factor. Alignment could be
done after this step. If dedicated functions are not provided by the system, suitable
scripts/macros for this task are freely available for software such as ImageJ,
although they may have been written for other techniques (e.g. Turboreg).
1.8 Reconstruction of an Aligned Tilt Series
As discussed in the preceding section, sequential micrographs acquired by slice and
view SEM are by definition already registered, and as such, represent xy slices of
the reconstructed volume. The remainder of this section applies to reconstruction of
tilt series acquired by TEM tomography. For a more detailed discussion, see
Chap. 5.
After achieving a satisfactory alignment (preferably according to some quantitative criterion), the practical execution of a reconstruction algorithm using modern
software packages can be as simple as a one-step operation using default settings.
This will generate a mathematically valid solution to the reconstruction problem,
which can be output as an xy image stack and/or animation. The defaults would
typically comprise a reconstruction of the entire volume corresponding to equal xyz
dimensions. Since a slab geometry is typically characterised by z dimensions that
are a small fraction of the xy area, reconstruction of equal xyz dimensions results in
1 Electron Tomography: A Primer
15
now implemented in some commercial as well as freely available software packages. The reader is referred to the review by Amat et al. [89] for an in-depth
discussion.
After alignment, reconstruction (see following section) is a comparatively simple
task that can be executed via command line, provided that the alignment meets
(user-defined) acceptable standards. For example, the operator might specify a given
tolerance corresponding to sub-pixel alignment. If the pixel size at the detector
corresponds to 1 nm, then sub-pixel alignment corresponds to sub-nanometre
accuracy. This should not be confused with resolution in the final reconstruction. It
simply means that if the alignment accuracy is better than 1 or 2 nm, then it will do
justice to the expected resolution of, say, 4 nm in the final reconstruction. In other
words, it does not degrade resolution.
By definition, the principle of SEM ‘slice and view’ tomography (Chap. 5)
would indicate that individual slices do not require further alignment: the block is
not tilted during acquisition, and the detector-block geometry is maintained
throughout the process, such that sequential images should be in perfect registration
as they are acquired. Specimen drift or other factors such as stage movement can
lead to a translational (xy) shift between one or more slices. These image shifts are
easily corrected during or after acquisition, however, and the process can be
automated. This is done most easily via a cross-correlation function, and it would
make sense to include this in the workflow, even if the unaligned stack passes
visual inspection. Typically, the first step in such a workflow would be to deal with
the massive file size via an initial xy scaling (‘binning’) factor. Alignment could be
done after this step. If dedicated functions are not provided by the system, suitable
scripts/macros for this task are freely available for software such as ImageJ,
although they may have been written for other techniques (e.g. Turboreg).
1.8 Reconstruction of an Aligned Tilt Series
As discussed in the preceding section, sequential micrographs acquired by slice and
view SEM are by definition already registered, and as such, represent xy slices of
the reconstructed volume. The remainder of this section applies to reconstruction of
tilt series acquired by TEM tomography. For a more detailed discussion, see
Chap. 5.
After achieving a satisfactory alignment (preferably according to some quantitative criterion), the practical execution of a reconstruction algorithm using modern
software packages can be as simple as a one-step operation using default settings.
This will generate a mathematically valid solution to the reconstruction problem,
which can be output as an xy image stack and/or animation. The defaults would
typically comprise a reconstruction of the entire volume corresponding to equal xyz
dimensions. Since a slab geometry is typically characterised by z dimensions that
are a small fraction of the xy area, reconstruction of equal xyz dimensions results in
1 Electron Tomography: A Primer
15