Extension of Field of View in CT Reconstruction

Using Elliptical Trajectory: A Simulation Study.

Mostafa Kamel, Jan Müller and Thorsten M. Buzug
Institute of Medical Engineering
University of Lübeck
Germany

{kamel, buzug}@ imt.uni-luebeck.de

Abstract
In this simulation experiment we investigate the effect an elliptical trajectory in a CT machine has on the field of view in
the reconstructed slices. A simulation of two scanning trajectories was carried out in order to obtain two different data-
sets: while the first one is a circular trajectory dataset with truncation, i.e. the object is larger than the field of view, the
second one is obtained with an elliptical trajectory using the same object and the same field of view size. A comparison
between the reconstructed slices from both datasets shows that the elliptical trajectory does extend the field of view.
However, there is a tradeoff between field of view and image quality.

1 Introduction In the experiment reported here, a CT simulation frame-

work has been set up to study the effects of an elliptical
In general, the field of view, which can be reconstructed trajectory on the reconstructed area. A simple extrapola-
from a CT scan, is limited by the detector size and the pro- tion for the missing data is done by extending the last
jection geometry (Fig. 1.1). Since the object or the patient known values to fill the new field of view.
are often larger than the field of view, this leads to prob-
lems in current CT scanners.
2 Simulation Framework
Scanning an object which is larger than the field of view In a simulation study four major factors must be inte-
will not only result in a truncated reconstruction but also grated, i.e. the X-ray source, the detector, the object and
in difficulties in reconstructing object parts, which is not the data sampling trajectory. In order to perform the simu-
consistently covered by the field of view (Fig. 1.1). Also, lation, these components are represented in software
without countermeasures, the truncation of the object will framework in the following simplified way:
often result in bright artifacts at the borders of the field of
view (Fig. 1.1). • A phantom is constructed by stacking a group of re-
constructed object slices to form a 3D matrix whose
elements contain the intensity values of a known re-
constructed object.
• The X-ray source is a monochromatic point source.
• The detector is a 2D matrix whose elements contain
the X-ray transform of the object.
• The trajectory is represented by the locus equations
that describe the curves of the X-ray orbit.

The CT simulation process is set up in two sub-processes.

Figure 1.1 The effect of truncated field of view on the re-
The first sub-process projects the volume points onto the
constructed image.
detector plane. The second sub-process moves the source-
detector system to the new scanning position. Hence, re-
The estimation of image data outside the field of view has peating the first process.
been discussed in many studies. For example, Ohnesorge
et al. (1) suggested a mathematical model to extrapolate In reality, X-ray radiation projects the volume by integrat-
for the missing information. Hsieh et al. (2) proposed a ing the true attenuation coefficients of the object on its
method for estimating the outside information depending path. In contrast, in the present simulation only grayscale
on the constant total attenuation of parallel geometry. values of points of the software phantom are used. These
Sourbelle et al. (3) introduced an adaptive method for intensity values are approximately proportional to the at-
extrapolating the image by estimating the convex hull of tenuation coefficients of points in the volume, so these are
the patient. All of these strategies are trying to remove the integrated instead.
truncation artifacts in circular trajectory CTs.
Since it is not possible to integrate infinite points along 3 Results
the rays (Fig. 2.1), only a finite number of points are cho-
sen. The selection of points is done by sampling the line
with a predefined sampling period (Fig. 2.1). A nearest- 3.1 Truncating the projections
neighbor interpolation is carried out in order to find the In the simulation study, two data sets are used: the first
attenuation values (intensity) at every point. Finally, the one is obtained by a simulation of a circular trajectory
values are summed up to form the projection value of the scanner for a head phantom, while the second set is ob-
ray. tained by a simulation of an elliptical trajectory scanner
for the same phantom (Fig. 3.1). In order to truncate the
datasets, a mask was applied which is smaller than the
actual detector (Fig. 3.1).

Figure 2.1 The interpolation points are calculated by

sampling the ray into points using a predefined sampling
period.

After each projection operation, , a new geometrical ar-

rangement for the X-ray source at its new position and the
object is calculated using both the initial scanning geome-
trical condition and the predefined locus equation for the
trajectory curve (Fig 2.2).

Figure 3.1 (a) the normal circular trajectory sonogram,

(b) elliptical trajectory sonogram, (c) truncated projec-
tions in circular trajectory and (d) truncated projections in
elliptical trajectory.
Figure 2.2 Upper view of the scanning system with two
curve examples, (left) a standard circular trajectory and
(right) an elliptical trajectory. 3.2 Extrapolating for the missing data
As mentioned before, the extrapolation method of the
Using the new geometry, new integration lines are missing data is a simple last neighbor method, which de-
formed, which leads to different points for interpolation. pends on the last known value. These values are extended
Hence, a new projection at a new position is formed (4). to fill the field of view (Fig. 3.2).

The validation of the simulation was accomplished by

creating a sequence of projections using a phantom made
from reconstructed slices without truncation. Then, the
reconstructed slices from the simulated sequence are vi-
sually compared with the original reconstruction
 4 Discussion and Conclusion
Examining the results shown in Fig. 3.3 demonstrates that
the elliptical trajectory is more appropriate for small de-
tector sizes. However, even if an object larger than scan-
ner’s field of view can be reconstructed, the quality of the
elliptically reconstructed slice is worse than the quality of
the circularly reconstructed slice (compared to the origi-
nal field of view). Consequently, a compromise between
the size of the field of view and the image quality must be
found.

With the described simulation framework, other trajecto-

ries are currently examined in order to find a configura-
tion, which both extends the field of view and preserves
the quality of the image. In addition, , the effect of other
Figure 3.2 (a) Extended sinogram for the elliptical trajec- filling methods on different trajectories are studied and
tory and (b) extended sinogram for the circular trajectory. evaluated.

5 References
3.3 Reconstructing the object
After extending the field of view using the last known [1] Ohnesorge, B, et al., Efficient correction for CT
value, a Feldkamp reconstruction algorithm is applied on image artifacts caused by objects extending outside
the extended set of data in order to obtain the recon- the scan field of view. 1, 2000, Med Physics, Vol.
structed slice (Fig. 3.3). 27, pp. 39-46.

[2] Hsieh, J, et al., A novel reconstruction algorithm to

extend the CT scan field-of-view. 9, 2004, Med
Physics, Vol. 31, pp. 2385-91.

[3] Sourbelle, K, et al., Reconstruction from truncated

projections in CT using adaptive detruncation. 5,
2005, European Radiology, Vol. 15, pp. 1008-1014.

[4] Mostafa, K, Simulation Framework for Cone-Beam

CT., Master Thesis, Luebeck of University, 2009.

Figure 3.3 (left-up) the reconstructed mid-slice from the

truncated projections in circular trajectory after extension,
(right-up) the reconstructed mid-slice of a complete se-
quence of projections without truncation, (left-bottom) the
reconstructed mid-slice from the truncated projections in
elliptical trajectory after extension and (right-bottom) the
elliptical truncated reconstruction compared with two
corners from the complete slice.