are generated by tilting the object to be reconstructed around an axis. Therefore, the
reconstruction process requires an accurate determination of the orientation of this
axis, which is common to all projections, to combine them in a correct way. In
addition, mathematical combination of projections needs that each one of the values
on the 3D numerical representation of the reconstructed object has been estimated
from data representing the same object at different tilt angles used during the
acquisition process. In case that the orientation of the tilt axis is not precisely
defined for each projection or that the values combined to compute the reconstruction do not correspond to an equivalent position in the original object, the
mixed information will result into inaccurate results and artifacts. Some classical
artifacts are blurred borders with small shifts or object deformation in “banana” or
“star” shapes (see Fig. 7.1). A detailed explanation of geometrical artifacts can be
found in [1].
In practice, as during the acquisition of the projection images the specimen is
placed in a holder, which is physically tilted by a goniometer inside the imaging
system, the precision and stability of motors to keep sample at the same exact
position during acquisition is unattainable at the nanometer scale expected for TET.
To compensate the lack of perfect tilting, automated acquisition tracks position and
focus of sample. However, despite the high performance achieved, this compensation is still not enough and gives rise to shifts and rotations. In addition, the
sample can be damaged during the acquisition process by suffering shrinkage or
simply the compensation process can slightly modify the position of the focal plane.
Fig. 7.1 Reconstructions of Pyrodictium abyssi. a XZ from a reconstruction with correct
alignment of tilt series images. b XZ from a reconstruction with incorrect alignment of tilt series
images. It can be seen that spherical shape are elongated to the right in a shape similar to banana
184
A. Verguet et al.
reconstruction process requires an accurate determination of the orientation of this
axis, which is common to all projections, to combine them in a correct way. In
addition, mathematical combination of projections needs that each one of the values
on the 3D numerical representation of the reconstructed object has been estimated
from data representing the same object at different tilt angles used during the
acquisition process. In case that the orientation of the tilt axis is not precisely
defined for each projection or that the values combined to compute the reconstruction do not correspond to an equivalent position in the original object, the
mixed information will result into inaccurate results and artifacts. Some classical
artifacts are blurred borders with small shifts or object deformation in “banana” or
“star” shapes (see Fig. 7.1). A detailed explanation of geometrical artifacts can be
found in [1].
In practice, as during the acquisition of the projection images the specimen is
placed in a holder, which is physically tilted by a goniometer inside the imaging
system, the precision and stability of motors to keep sample at the same exact
position during acquisition is unattainable at the nanometer scale expected for TET.
To compensate the lack of perfect tilting, automated acquisition tracks position and
focus of sample. However, despite the high performance achieved, this compensation is still not enough and gives rise to shifts and rotations. In addition, the
sample can be damaged during the acquisition process by suffering shrinkage or
simply the compensation process can slightly modify the position of the focal plane.
Fig. 7.1 Reconstructions of Pyrodictium abyssi. a XZ from a reconstruction with correct
alignment of tilt series images. b XZ from a reconstruction with incorrect alignment of tilt series
images. It can be seen that spherical shape are elongated to the right in a shape similar to banana
184
A. Verguet et al.