1

IMAGE QUALITY IN EXTENDED ARC FILTERED DIGITAL

TOMOSYNTHESIS

C. Badea, Z. Kolitsi and N. Pallikarakis

Department of Medical Physics, University of Patras, Patras 26500, Greece

Correspondence: Nicolas Pallikarakis, Department of Medical Physics,

University of Patras, Patras 26500, Greece

Tel: +30-61-996112

Fax:+30-61-999855

e-mail: nipa@bme.med.upatras.gr

2

Abstract

Purpose: To study image quality in Filtered Digital Tomosynthesis (FDTS)

tomograms as a function of their reconstruction arc, using isocentrically

acquired, fluoroscopic projection data.

Material and Methods: Both Digital Tomosynthesis (DTS) and Cone Beam CT

(CBCT) reconstruction algorithms are based on backprojection and use cone

beam projection data as input. Under limited angle conditions, CBCT is reduced

to FDTS, where only a subset of projection data used for reconstruction.

The effect of the reconstruction arc on the spatial resolution, slice thickness,

contrast sensitivity, shape distortion and artifacts, was also experimentally

studied. The investigation was performed using both simulated and actual

fluoroscopic images.

Results and Conclusion: Image quality in terms of spatial resolution, slice

thickness, shape distortion and artifacts, improved with increasing

reconstruction arc and was optimized at 1800, while contrast continued to

improve as the arc was increased to 3600. However DTS was assessed to be

the technique of choice when reconstruction arcs of less than 40 degrees are

used. Consequently, FDTS may be successfully implemented in applications

involving extended arc reconstructions, in the range between 400 delimiting the

DTS domain and 3600 corresponding to CBCT.

Key words: tomosynthesis digital; imaging; volume; angiography.

3

Cone Beam CT (CBCT) has been used with radiotherapy simulators, to produce

therapeutic-quality images, as those obtained using the systems’ CT option,

which satisfy the data specification for 3D treatment planning software (2).

CBCT exploits cone beam projection data acquired over a complete arc by

using a planar detector such as an image intensifier (3). There are, however,

situations where collecting projection data over 3600 is not feasible; in such

cases, limited angle reconstruction methods may be used.

DTS is a method of limited angle reconstruction of tomographic images,

produced at variable heights, on the basis of a set of angular projections (5).

The DTS reconstruction algorithms simulate classical tomography by either

translating and adding the projection images (8) or by deploying the

backprojection method, to transfer and deposit data at the intersection of each

ray with the plane of reconstruction (6). In all approaches, a limitation in the size

of sampling arc arises from the restricted movement of the tube-detector

assembly. DTS tomograms are invariably affected by tomographic noise, i.e.

blurred images of structural detail, lying outside the plane of interest, and

superimposed over the focused image of the fulcrum plane. Several

contributions have focused on enhancing the quality of these images by a

variety of methods involving post-processing of tomograms (1, 9), or by pre-

processing of projections (7, 11). Filtered Digital Tomosynthesis (FDTS)

involves the application of filtering to the projection data, followed by

backprojection (11).

This article comprises a study of image quality for FDTS reconstructed

tomograms using data collected with isocentric fluoroscopic units, such as the

radiotherapy simulator or the angiographic units. The effect of the reconstruction

arc on spatial resolution, slice thickness, contrast sensitivity, shape distortion

and artifacts, was experimentally investigated.

Material and Methods

Both DTS and CBCT reconstruction algorithms are based on backprojection and

use cone beam projection data as input. Under limited angle conditions, CBCT

4

is reduced to FDTS, with only a subset of projection data used for

reconstruction.

The overall effect of filtering of projection images as well as the variation of a

basic set of image quality parameters as a function of the reconstruction arc

was investigated using simulated projection data and geometries of the

radiotherapy simulator. Simulated projections were generated with the use of a

software data generator for radiographic imaging investigations (12).

Additionally, a subjective visual assessment of image quality under different

reconstruction arcs, using DTS and FDTS was performed on tomograms

obtained using actual angiographic projection data from a contrast-enhanced

sheep lung. The phantom was imaged with the Philips DVI, Digital Angiographic

unit, operated at a source to isocenter distance equal to 67.5 cm and a source

to image intensifier distance equal to 92.5 cm. The imaging chain had a limiting

resolution of 12 lp/cm. Projection images were acquired every 20 over a 1200

arc using a DTS prototype system (10). All projections were filtered along each

row with a standard ramp filter and a Hamming window.

The following image characteristics were studied:

(i) Spatial resolution

Spatial resolution was measured using a simulated bar pattern phantom with

line pairs varying in size from 1 to 20 lp/cm. The object was considered to have

a thickness of 1mm and to be inclined at 450 with respect to the source detector

rotation plane. The oblique orientation of the bars was chosen in order to

provide visual information on the system’s ability to filter out noise at various

distances from the coronal plane of the phantom (13). The source to isocenter

distance was set to 100 cm and the source to image intensifier distance to 130

cm. The imaging chain was considered to have a limiting resolution of 20 lp/cm.

Projection images were generated at every 20 over a 3600 arc. Both DTS and

FDTS tomograms were reconstructed at various reconstruction arcs. Spatial

resolution was assessed by means of the Square Wave Response Function

(SWRF), determined over the complete range of spatial frequencies.

5

(ii) Slice Thickness

Slice thickness was measured using the same set of data described above.

Measurements were performed for each spatial frequency by means of the Full

Width at Half Maximum (FWHM) of the intensity profiles taken along the

transverse axis of the corresponding pair of bars (4).

(iii) Contrast sensitivity

A second electronic phantom was designed for contrast sensitivity

investigations, consisting of an ellipsoid containing six equally sized spheres

(R=0.5 cm), equidistantly spaced and with their centers along the long diameter

of the middle coronal slice of the phantom. The spheres were of varying

densities ranging from 0 to 2.00 g/cm3 providing background contrast variations

of 2%, 5%, 10%, 20%, 30% and 100%. Contrast sensitivity, defined as the

difference in signal strength between the structure of interest and the

background was studied as a function of the reconstruction arc. For each

sphere, the arc needed to establish a minimum 2% contrast relative to the

background, was determined.

(iv) Shape distortion and artifacts

Artifacts and shape distortions are present in limited angle reconstructions. In

order to assess them, a spherical phantom was used, composed of a high

density material, placed inside a homogeneous medium of unit density. The

phantom was imaged with its center positioned on the isocenter. Central axial

tomograms, coinciding with the sampling plane, were reconstructed for different

reconstruction arcs and the ratios of the FWHM of the diameter profiles along

the horizontal and vertical directions were used for assessment of shape

distortions.

Results

Fig. 1 shows the SWRF plots using both the DTS and FDTS tomograms. The

slice thickness variation with the size of reconstruction arc for both DTS and

FDTS are presented in Fig. 2. Studies for a high (5 lp/cm) and a low (1 lp/cm)

spatial frequency are shown on the same graph. Contrast sensitivity for a range

6

of background contrast conditions is given in the table. Shape distortion

dependency on the reconstruction arc is shown in Fig. 3. The effect of

increasing reconstruction arc on images of anatomic content is shown in Fig. 4,

for reconstructed tomograms of the contrast-injected sheep-lung phantom.

Three parallel coronal planes at 5 mm spacing distance were reconstructed with

DTS and FDTS using reconstruction arcs of 400 and 1200, respectively.

Discussion

The extent of the reconstruction arc has a significant impact on spatial

resolution (Fig.1). In the higher frequency range the two techniques provide

comparable results for arcs smaller than 1200. However, for larger

reconstruction arcs, FDTS demonstrates a clear advantage. On the other hand,

DTS may be considered the technique of choice when reconstruction arcs in the

range of 400 or less are used. An increase in the arc results in tomograms of

finer slice thickness (Fig.2). As shown by the table, high contrast structures,

such as vascular structures or markers used in radiotherapy, may be detected

under limited angle conditions, while low contrast structures will appear only on

extended arc reconstructions. In fact, our results show that, a full 3600

reconstruction i.e., CBCT, is necessary to detect a 2% contrast variation to the

background. The shape distortions present in limited arc reconstructions,

introduce problems (axial elongations) when defining contours of otherwise

discernable structures. Again, as seen in Fig. 3, reconstruction arcs of 1800 or

larger are required to eliminate such shape distortions. The same conclusion

can be drawn from the slice thickness results, shown in Fig. 2, not surprisingly,

since both concepts are related to the filtering of out-of-plane information.

Consequently, image quality in terms of, high contrast spatial resolution, slice

thickness, shape distortion and artifacts, improves with the reconstruction arc

and is optimized at 1800, while contrast continues to improve when the arc is

increasing towards 3600.

Conclusions

7

Based on the above observations, it may be stated that the sampling range

between 400 and 3600, delimited by DTS and CBCT respectively, is the domain

where FDTS can be applied to the advantage of applications where sampling

over an extended arc is possible. FDTS with an extended arc may even

successfully substitute CBCT in applications, where densitometric information is

not required.

8

References

1) BADEA C., KOLITSI Z. & PALLIKARAKIS N.: A wavelet-based method for removal

of out-of-plane structures in digital tomosynthesis. Comput Med Imaging

Graph. 22 (1998), 309.

2) CHO P., JOHNSON R. & GRIFFIN T.: Cone-beam CT for radiotherapy

applications. Phys. Med. Biol. 40 (1995), 1863.

3) FELDKAMP L., DAVIS L. & KRESS J.: Practical cone-beam algorithm. J. Opt.

Soc. Am. A1 (1984), 612.

4) GOODENOUGH D.J., LEVY J.R. & KASALES C.: Development of phantoms for

spiral CT. Comput. Med. Imaging Graph. 22 (1998), 247.

5) GRANT D.G.: Tomosynthesis: a three-dimensional radiographic imaging

technique. IEEE Trans. Biomed. Eng. 19 (1972), 20.

6) KAMPP T.D.:The Backprojection Method Applied to Classical Tomography.

Med. Phys. 13 (1985), 329.

7) KNUTSSON H.E., EDHOLM P., GRANLUND H.G. & PETTERSON C.U.:

Ectomography ¯¯ a new radiographic reconstruction method. I. Theory and

error estimates. IEEE Trans. Biomed. Eng. 27 (1980), 641.

8) KOLITSI Z., PANAYIOTAKIS G., ANASTASSOPOULOS V., SKODRAS A., &

PALLIKARAKIS N.: A multiple projection algorithm for digital tomosynthesis.

Med. Phys. 19 (1992), 1045.

9

9) KOLITSI Z., PANAYIOTAKIS G. & PALLIKARAKIS N.: A method for selective

removal of out-of-plane structures in digital tomosynthesis. Med. Phys. 20

(1993), 47.

10) KOLITSI Z., YOLDASSIS N., SIOZOS T. & PALLIKARAKIS N.: Volume Imaging in

fluoroscopy: a clinical prototype system based on a generalised digital

tomosynthesis technique. Acta Radiol. 37 (1996), 741.

11) LAURITSCH G., HAERER W.: Theoretical framework for filtered back projection

in tomosynthesis. In Proceedings SPIE Conference In Image Processing

3338 San Diego (1998),1127.

12) LAZOS D., KOLITSI Z. & PALLIKARAKIS N.: A software data generator for

radiographic imaging investigations. IEEE Trans Inf. Technol. Biomed. 4(1),

(2000), 76.

13) PETERSSON C.U.: Computer-simulated object for testing principles for

radiographic reproduction of slices, applied to ectomography. Med. Biol.

Eng. Comput. 20 (1982), 425.

10

List of Acronyms

DTS Digital Tomosynthesis

FDTS Filtered Digital Tomosynthesis

CBCT Cone Beam CT

FWHM Full Width at Half Maximum

SWRF Square Wave Response Function

11

FIGURE CAPTIONS

Fig. 1. SWRF plots for different reconstruction arcs, for (a) DTS and (b) FDTS

reconstructed images. Simulated projection data for a bar pattern phantom have

been used.

Fig. 2. Slice thickness variation in DTS and FDTS with the size of reconstruction

arc. The effect has been studied for a high (5 lp/cm) and a low (1 lp/cm) spatial

frequency.

Fig. 3. Shape distortion and artifacts: the center axial section of the sphere has

been reconstructed and the ratio of the FWHM of the diameter profiles along the

horizontal and the vertical directions was plotted as a function of the

reconstruction arc. An axial tomogram reconstructed using an arc of 600 is

shown in the insert.

Fig. 4. A tomographic sequence of three transversal planes distanced at 0.5 cm

through the angiographic phantom reconstructed using DTS and FDTS at 400

and 1200 reconstruction arcs.

Table. Minimum reconstruction arc required to establish a 2% contrast

detectability for different background contrast.

12

0 2 4 6 8 10 12 14 16 18 200.0

0.2

0.4

0.6

0.8

1.0

DTS 40 0

DTS 80 0

DTS 120 0

DTS 180 0

DTS 360 0

SW

RF

Spatial Frequency (lp/cm )

0 2 4 6 8 10 12 14 16 18 200.0

0.2

0.4

0.6

0.8

1.0

FDTS 40 0

FDTS 80 0

FDTS 120 0

FDTS 180 0

FDTS 360 0

SW

RF

Spatial Frequency(lp/cm )

(b)

(a)

13

0 20 40 60 80 100 120 140 160 180

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

Slic

e Th

ickn

ess

(cm

)

Reconstruction arc (degrees)

DTS (1 lp/cm)

DTS (5 lp/cm)

FDTS (1 lp/cm)

FDTS (5 lp/cm)

14

0 50 100 150 200 250 300 3500.0

0.2

0.4

0.6

0.8

1.0

FWH

M H

OR

IZO

NTA

L / F

WH

M V

ER

TIC

AL

RECONSTRUCTION ARC (degrees)

15

DTS 400

DTS 1200

FDTS 400

FDTS 1200

Z=1.5cm Z=2cm Z=2.5cm

16

Object 1 2 3 4 5 6

Object Contrast 2% 5% 10% 20% 30% 100% Min Arc for contrast detectability 3600 1440 700 360 200 00

17