29.11.2012 Views

EPID IMAGE QUALITY: LAS VEGAS PHANTOM AS AN OBJECTIVE ...

EPID IMAGE QUALITY: LAS VEGAS PHANTOM AS AN OBJECTIVE ...

EPID IMAGE QUALITY: LAS VEGAS PHANTOM AS AN OBJECTIVE ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Witold Skrzyński 1<br />

Andrew J. Reilly 2<br />

David I. Thwaites 2<br />

Wojciech Bulski 1<br />

1 Centre of Oncology, Warsaw, Poland<br />

2 Edinburgh Cancer Centre, Western General Hospital, Edinburgh, UK<br />

w.skrzynski@rth.coi.waw.pl<br />

.<br />

<strong>EPID</strong> <strong>IMAGE</strong> <strong>QUALITY</strong>: <strong>L<strong>AS</strong></strong> <strong>VEG<strong>AS</strong></strong> <strong>PH<strong>AN</strong>TOM</strong><br />

<strong>AS</strong> <strong>AN</strong> <strong>OBJECTIVE</strong> TOOL<br />

Abstract: This work is a comparison of three approaches for<br />

<strong>EPID</strong> image quality control using the Las Vegas phantom:<br />

contrast-detail counting by human observers, an automatic<br />

detail counting technique and the direct calculation of performance<br />

indices (SNR and MTF). The standard detail<br />

counting procedure is observer-dependant and insensitive to<br />

artifacts, whilst the automatic method is robust and fully<br />

objective. Results of MTF calculations were in excellent<br />

agreement with those obtained using a commercial QC-3V<br />

phantom. Basic image analysis software thus enables the<br />

Las Vegas phantom to be used as an objective tool for<br />

evaluating the quality of electronic portal images.<br />

1. INTRODUCTION<br />

<strong>EPID</strong>s (Electronic Portal Imaging Devices) are<br />

widely used for the assessment of geometrical accuracy<br />

in modern radiotherapy [1]. Portal images of a high and<br />

consistent quality are essential for this task.<br />

The Las Vegas phantom (Figure 1) is a contrast-detail<br />

phantom supplied with most <strong>EPID</strong>s for acceptance<br />

testing [2] and routine image quality control. The<br />

control procedure suggested by manufacturers is usually<br />

limited to a simple check of the number of details visible<br />

in the image by a human observer.<br />

Increasing contrast<br />

(hole depth)<br />

Increasing detail size<br />

(hole diameter)<br />

Fig. 1. Schematic view of Las Vegas phantom<br />

Several other phantoms and test methods are also in<br />

common use. One of the most popular is the commercial<br />

QC-3V phantom with high-contrast bar resolution patterns<br />

and uniform regions of interest [3]. This is supplied<br />

with the PIPSpro [4] software for automated analysis of<br />

image quality. PIPSpro calculates the MTF (modulation<br />

transfer function) and SNR (signal to noise ratio). Recently<br />

it has been suggested that a very simple phantom<br />

and home-made software might be used to objectively<br />

measure image quality parameters [5].<br />

This work considers two objective approaches for<br />

image quality control with the Las Vegas phantom: detail<br />

Warsaw, 29-30 Sept.<br />

2005<br />

counting by a computer observer, and utilizing the phantom<br />

in a new mode to calculate SNR and MTF.<br />

2. MATERIALS <strong>AN</strong>D METHODS<br />

2.1 <strong>EPID</strong>s<br />

Measurements were performed using Varian PortalVision<br />

aS500 (amorphous silicon) <strong>EPID</strong>s in Edinburgh<br />

Cancer Centre and aS500 and LC250 (liquid ion<br />

chamber) <strong>EPID</strong>s installed in the Centre of Oncology in<br />

Warsaw. All of the <strong>EPID</strong>s were installed on Varian Clinac<br />

linear accelerators.<br />

2.2 Software<br />

All the images were analyzed using the ImageJ<br />

software platform [6]. Several complex image analysis<br />

routines were required for this work and custom ImageJ<br />

‘plugins’ were developed to implement these.<br />

2.3 Detail counting<br />

Las Vegas phantom images for five aS500 <strong>EPID</strong>s in<br />

Edinburgh Cancer Centre were acquired on a weekly<br />

basis over a period of 9 months. For each image, the<br />

number of visible details was counted immediately following<br />

acquisition. Three trained observers performed<br />

the task on a rotational basis and the results have been<br />

investigated for intra- and inter-observer variability.<br />

A statistical algorithm for automated detail counting<br />

([7], modified) was also implemented. Results obtained<br />

by human and computer observers were then compared.<br />

2.4 MTF and SNR calculation<br />

Pairs of images of Las Vegas and QC-3V phantoms<br />

were acquired for various beam energies and dose rates.<br />

For the Las Vegas phantom the MTF was calculated<br />

as the magnitude of the Fourier transform of the<br />

edge spread function (ESF) [8]. The difference between<br />

the mean pixel value in a uniform region of the phantom<br />

and that of a region outside the phantom was taken as the<br />

signal for the SNR calculation.<br />

Images of the QC-3V phantom were first analyzed<br />

with the PIPSpro software, which outputs MTF normalized<br />

to unity at the frequency of 0.1 line-pairs per mm<br />

(lowest frequency bar test pattern). These were then reanalyzed<br />

using a modified method [9] which outputs the<br />

“correct” MTF (normalized to unity at zero frequency).


3. RESULTS <strong>AN</strong>D DISCUSSION<br />

3.1 Detail counting<br />

In Figure 2 the performance distribution of two of<br />

the Edinburgh observers is plotted and illustrates interobserver<br />

variability. Figure 3 presents a comparison of<br />

results obtained by human and computer observers for<br />

one <strong>EPID</strong>. Results obtained by human observers appear<br />

to show that image quality is improving over time. This<br />

positive trend is rather unexpected (detectors are ageing)<br />

and is thought to be as a result of observers becoming<br />

progressively more confident in identifying details.<br />

Computer analysis of the images suggests that image<br />

quality is stable over time.<br />

The number of human or computer-detected details<br />

was not significantly affected by image artifacts (e.g.<br />

horizontal banding, defective pixels). Detail-counting is<br />

therefore is not a good measure of overall image quality.<br />

number of counts<br />

30<br />

20<br />

10<br />

observer1<br />

observer2<br />

0<br />

0,9 1,0 1,1<br />

Las Vegas test result<br />

(number of visible holes normalized to mean)<br />

Fig. 2. Number of visible structures in images<br />

of Las Vegas phantom for two observers<br />

number of visible holes<br />

24<br />

22<br />

20<br />

0 10 20 30 40<br />

24<br />

22<br />

20<br />

human observer(s)<br />

computer observer<br />

0 10 20<br />

week<br />

30 40<br />

Fig. 3. Las Vegas phantom test results for human and<br />

computer observers<br />

3.2 Modulation Transfer Function<br />

Typical results of the MTF calculations are presented<br />

in Figure 4. The curves obtained from the analyses<br />

of the edge profile (Las Vegas phantom) and bar test<br />

patterns (QC-3V phantom) are in excellent agreement.<br />

As expected, the results obtained with PIPSpro do not<br />

agree with other methods due to non-standard MTF normalization.<br />

(The PIPSpro curve is shifted to the right.)<br />

For routine quality control purposes the SNR and<br />

the frequency at which the MTF is 50% (f50) are typically<br />

recorded and it was found that these quantities are sensitive<br />

to image artifacts. They are therefore a useful indicator<br />

of when technical maintenance may be required.<br />

MTF<br />

1,0<br />

0,5<br />

0,0<br />

0,0 0,5 1,0<br />

Las Vegas - edge profile<br />

QC-3V (PIPSpro results)<br />

QC-3V (corrected MTF)<br />

resolution (line pairs per mm)<br />

Fig. 4. MTF obtained with three methods<br />

4. CONCLUSIONS<br />

The standard detail counting procedure is observerdependant<br />

and insensitive to artifacts. However, basic<br />

analysis software enables the Las Vegas phantom to be<br />

used as a fully objective tool which is sensitive to artifacts<br />

and hence suitable for inclusion in a quality control<br />

program.<br />

LITERATURE<br />

[1] P. Munro, ‘‘Portal imaging technology: Past, present and<br />

future,’’ Semin. Radiat. Oncol. 5, 115–133, 1995<br />

[2] Low D.A., Klein E.E., Maag D.K., Umfleet W.E., Purdy<br />

J.A., Commissioning and periodic quality assurance of a<br />

clinical electronic portal imaging device, Int. J. Radiation<br />

Oncology Biol. Phys. 34(1), p.117, 1996<br />

[3] Shalev S., Test phantoms for <strong>EPID</strong>s: A comparison<br />

between the Las Vegas and the QC-3 phantoms, Masthead<br />

Imaging Corporation, 2003<br />

[4] Shalev S., PIPSpro, version 3.2.2, Masthead Imaging<br />

Corporation, British Columbia, Canada, 2002<br />

[5] Reilly A.J., Skrzyński W., Thwaites D.I., <strong>EPID</strong> Quality<br />

Assurance: A Fresh Approach, Clinical Oncology<br />

17(supp. 1), p.219, 2005<br />

[6] Rasband W.S., ImageJ, U. S. National Institutes of<br />

Health, Bethesda, Maryland, USA,<br />

http://rsb.info.nih.gov/ij/, 1997-2005.<br />

[7] Dong L, Boyer A.L., An objective method for evaluating<br />

electronic portal imaging devices, Med. Phys. 21(6),<br />

p.755, 1994<br />

[8] International Electrotechnical Commission, International<br />

Standard IEC 62220-1: Medical electrical equipment –<br />

Characteristics of digital imaging devices – Part 1: Determination<br />

of the detective quantum efficiency, Geneva,<br />

2003<br />

[9] Droege R.T., Morin R.L., A practical method to measure<br />

the MTF of CT scanner, Med. Phys. 9(5):758-760, 1982<br />

This work has been supported by the IAEA (Fellowship No.<br />

C6/POL/03016). The authors gratefully acknowledge the<br />

contribution of L Cherry, A Doig and W Ronaldson of<br />

Edinburgh Cancer Centre in analyzing the Las Vegas Phantom<br />

images.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!