More Info
P Sharp, University of Aberdeen

One of the problems facing the physicist is that while there is very strong support for the
value of measuring image quality and no shortage of proposals on how it might be
measured, there is little agreement on the best way to do it. Image quality can be addressed
by considering either the quality of the data acquired by the imaging instrument or that of the
displayed image. The former is most familiar in addressing the problems of quality control of
imaging instrumentation, as in the use of objective measures of quality such as modulation
transfer function. The latter requires the use of the human observer to view some test
pattern and pass a subjective judgement on what can be seen. The former has the
advantage of being an objective and perhaps impartial measure of quality but suffers from
the major drawback of simply measuring one specific aspect of device performance. Viewing
the displayed data approach offers the superficially attractive advantage of using the real
image, the end product of the imaging process, on which to base an assessment of quality.
The disadvantage is that measurements with human observers can be confounded by the
variability in their performance.

In Report 54 the ICRU set out the background to a unified approach to the measurement of
image quality incorporating both the objective and subjective approaches under one
framework. This document was seen as a generic approach and reports dealing with the
application to specific imaging modalities are in preparation..

Session 1                                     1
R Bury, Leeds General Infirmary

The meeting will include many references to the objective measurement of image quality,
and rightly so: if you can't measure it, it's difficult to have any sensible discussion about the
effects of dose reduction. Unfortunately, when radiologists talk about image quality, they are
often really talking more about the clinical adequacy of the image, and although most of us
think we know what we mean by that, the definition of ‘       adequate’is elusive. It is a simple
matter to plot objective measures of image quality against dose - only two axes are required.
Clinical adequacy requires (at least) one additional axis; that of justification.

And the choice of an outcome measure that will act as an indicator of image adequacy is
also difficult. Clearly, the ability to make a diagnosis from the image is a vital element, but
there are different degrees of certainty in the drawing of diagnostic conclusions from a film,
and so ROC curves are likely to be putting in an appearance at some stage. And once you
are talking ROC analysis and, probably, phantom studies, you begin to move away from
clinical practice into a more artificial laboratory situation. These issues will all be discussed,
and I shall review the literature. I shall also refer to some of the work we have done at Leeds
which begins to address these problems.

In 1998 the few ideas we did have on this topic are being challenged by the arrival of digital
imaging and soft copy reporting - radiological images are beginning to look very different,
and the possibilities for image processing and presentation are increasing almost daily. It is
therefore even more important that we develop robust methods of measuring imaging
performance in the clinical workplace.

Session 1                                       2
C W Robertson

The Medical Device Directive 93/42/EEC requires manufacturers to affix the CE mark,
indicating full compliance with its provisions. It contains a requirement for diagnostic X-ray
equipment that it “  shall be designed and manufactured in such a way as to achieve
appropriate image and/or output quality for the intended medical purpose whilst minimising
radiation exposure of the patient and user” An objective international standard to address
this requirement is being prepared by a working group of the International Electrotechnical
Commission (IEC). The first draft will shortly be circulated to the participating countries. The
draft standard proposes to require manufacturers to declare "representative uses" from a
list. For each representative use tests are specified in which the attainment of relevant
image properties is required to be demonstrated without exceeding specified radiation
levels. In this paper, a general description is given of the proposed range and classification
of the representative uses and of the nature of the tests. Examples of compliance criteria are
included. The tests are designed for product type certification at the point of manufacture
and not for use after delivery. When finalised, the standard will take its place alongside
others providing the user with assurance of the safety and effectiveness of X-ray equipment
for its intended uses. It is likely that the standard will require the user to be informed of the
representative uses for which compliance has been determined, but it will contain no
requirements to limit or regulate the clinical use of equipment.

Session 1                                      3
C J Martin, Health Physics, Department of Clinical Physics and Bio-Engineering,
West Glasgow Hospitals University NHS Trust

In order to obtain images which are adequate for the clinical purpose with the minimum
radiation dose to the patient, methods are required to evaluate the performance of imaging
systems. Measurements using test objects alone describe the behaviour of equipment under
specific conditions and usually provide a measure of performance of the image receptor
under ideal conditions. They do not take account of variations in radiation quality and
scattered radiation, which are factors likely to be changed in optimising the balance between
radiation dose and image quality for the whole system, nor can they be linked readily to
clinical imaging performance.

A full evaluation of image performance can only be obtained from lengthy trials to determine
clinical outcomes, but these methods are not suitable for routine use in hospitals. Diagnostic
performance criteria referring to visualisation of anatomical features in normal individuals
are being established for simple assessments of clinical images. Scoring of such quality
criteria for radiographic and CT images provides a simple method through which image
quality can be assessed in every hospital department to ensure their adequacy for
diagnosis. Imaging of a patient can be simulated by incorporating a test object in a tissue
equivalent phantom and provides a technique which physicists can use in assessing
performance of the whole imaging system, but for the test objects which most of us use, little
data is available on contrast with X-ray spectra transmitted through tissue. There is a wide
range of approaches to measurement of image quality. The techniques which are simplest to
use rely on subjective assessments and do not provide a true objective measure of
performance, but are they sufficient for general hospital use?. How much should we be
doing to assess image quality in hospital departments and which techniques should we use?

Session 1                                     4
FLUOROSCOPIC    IMAGE    QUALITY:   USE                                                 OF
D J Gallacher, A Mackenzie, S Batchelor, J Lynch, J Saunders, Medical Physics
Directorate, Guy's and St. Thomas' Hospitals Trust, Lambeth Palace Road,
London SE1 7EH

The use of threshold contrast detail detection test objects in image quality testing is
widespread for fluoroscopic x-ray systems. Various reference data on expected levels of
image quality performance are available. In this study the use of standard reference
threshold detection index curves has been investigated with a view to establishing an image
quality index which can be used to assess serial testing performance of x-ray TV
fluoroscopy systems. A comparison has been made between different fields sizes and input
kerma rates at the intensifier to establish from standard reference data the self consistency
of this approach. Serial test data under standard test conditions from several systems is
presented to demonstrate the validity of this approach. Also comparison is made between
results obtained under variable distance viewing protocols and the established fixed
distance image quality index. A single image quality index has been shown to be a useful
parameter in assessing system performance in conjunction with other system test data.

Session 1                                    5
I A J Fife & C P Lawinski, KCARE, King’ College Hospital, East Dulwich Grove,
Dulwich, London SE22 8PT

Apart from ensuring that the systems evaluated conform to the appropriate requirements on
electrical, mechanical and radiation safety, a series of additional tests were carried out to
examine image quality and patient dose. The imaging performance was evaluated using
typical threshold detectability test objects as well as an objective assessment of limiting
resolution and focal spot size. Patient dose together and image intensifier entrance dose
were also measured. As the technology of imaging equipment advances, the specification
of system improves and the image quality/patient dose ratio is expected to increase.

Additionally, the evaluation includes and appraisal of ease of use during clinical
examinations which is valuable in highlighting user problems. This type of assessment and
comparison is most useful when purchasing equipment where fitness of purpose is important
in the overall cost analysis. It is common for over specified systems to be purchased which
are often more difficult to use where versatility tends to be lost with system complexity.

The results suggest that the image intensifier entrance dose levels are now lower than the
typical levels quoted in IPEM Report 32. Additionally, the measured entrance surface doses
were in all cases less than the maximum level given in the Guidance Notes. In terms of
image quality, the modern systems did not show a significant improvement with respect to
older reference systems.

Session 1                                    6

G Weatherburn, Health Economics Research Group, Brunel University,
Uxbridge, Middlesex UB8 3PH

This study was undertaken as part of an evaluation of a small Picture Archive and
Communications System (PACS) linking the Radiology Department and the Intensive
Therapy Unit (ITU) at Glan Clwyd Hospital and was financed by NHS Wales. In this hospital,
the ITU clinicians viewed soft copy images in ITU and the radiologists reported the same
examinations from hard copy images in Radiology. Conventional film/screen images were
used throughout the rest of the hospital. Thus, this study aimed to compare these three
types of images using standard test tools. Leeds test objects were used to compare film, CR
hard copy and PACS images in terms of high contrast resolution, low contrast detectability,
effective dynamic range and threshold contrast detail detectability. In addition, for threshold
detail detectability, the effect of increase in mAs on all three types of images was
investigated. All images were viewed and scored by four medical physicists. The results of
this study will be presented.

Session 1                                     7
C L Skinner, I A Castellano Smith & S H Evans, Physics Department, Royal
Marsden NHS Trust, Fulham Road, London SW3 6JJ

Computed radiography (CR) using phosphor plates was first introduced in 1981. However, it
appears that there is still a long way to go before the installation of a CR system and its
assimilation into clinical use becomes a well-defined, routine exercise. We have recently
been involved with such an installation and experienced some difficulty in obtaining certain
information we required concerning the CR system from the installers. The manufacturers of
the X-ray equipment involved were largely unaware of the new requirements on their units,
particularly with regard to kV compensation under automatic exposure control and input
dose to the phosphor plate. A thorough revision of film-screen radiography techniques was
required which presented some difficulties to the users. We believe that these issues need
to be addressed before clinical optimisation can be undertaken in the digital X-ray
department. Potential purchasers should be aware that doses required for computed
radiography can be significantly higher than those for a standard speed film-screen system.

Session 1                                    8
D Sutton, Medical Physics Department, Ninewells Hospital, Dundee

Radiation quality influences image quality and radiation dose through the mechanisms by
which photons of different energy interact with the tissue. Tube potential determines the
proportion of high energy photons in an X-ray beam. A theoretical example will be
presented in order to demonstrate the influence of tube potential on both entrance surface
dose and effective dose for abdominal antero-posterior (AP) and postero-anterior (PA)
radiographs. An abdomen is represented by a 150 mm thick layer of tissue and 40 mm of
bone and theoretical X-ray spectra were derived for tube potentials between 60 and 120
kVp. The effect of filtration will be demonstrated using the same theoretical model. In this
case, X-ray spectra have been generated in order to show differences between the ‘    input
spectra’incident on a patient’ skin and the ‘ output spectra’which would be incident on the
image receptor and therefore contribute to image formation. Filters such as copper and
erbium are considered and potential dose reductions presented.

Click here to see the slides used at this presentation.

Session 2                                      9
THE EFFECT                  OF 0.1mm ADDITIONAL COPPER-
W A Hummel & R J Zijlstra, KCL & MCL Leeuwarden, Netherlands

It is well known that the patient (entrance) dose can be reduced by adding extra filtration to
the X-ray beam. This additional filtration however influences the X-ray spectrum which
effects the image contrast. In order to determine the effects of added copper-filtration the
operating parameters are kept constant. Therefore we used a fixed combination between
phantom thickness and tube potential.

The used combinations were based on the clinical practice of our hospital (see figure 1).

The study contains three parts: A) measurement of the entrance dose, B) comparison of the
physical image quality and C) clinical evaluation of X-ray images of hands or feet.

A) Entrance dose.
   For 11 different phantom thickness/tube potential combinations the entrance dose has
   been measured with and without 0.1 mm added copper-filtration.

   Result: An overall entrance dose reduction of (42 + 7)%

B) Scoring of the image quality.
   For 6 different phantom thickness/tube potential combinations 1 cm of the perspex-
   phantom was replaced by a Contrast-Detail or Burger-Rose phantom.
   In our hospital 2 types of film/screen combinations1 are used. For X-ray images of hands
   or feet from patients with rheumatoid arthritis a low speed class system is used. All the
   other images are made on the high speed class system.
   For every combination two images were made, one with and one without added copper-
   filtration. From the judgement of a panel of 10 observers a quality measure was obtained
   for the X-ray images.

   Result: The images made with the 0.1 mm added copper-filtration on the low speed class
   film/screen combination showed a significant loss of contrast.
   On the other film/screen combination the loss in contrast was smaller.

Session 2                                    10
C) Clinical evaluation.
   For 33 patients apart from two the normal X-ray image of the hand, finger or foot, an
   additional image was made with 0.1 mm added copper filtration.
   Both images were scored by radiologists on too light or too dark, perceptibility of small
   details, ability of giving a diagnosis.

   Preliminary result: Low speed class film/screen combination: an unacceptable loss of the
   perceptibility of small details as well as the ability of giving a diagnosis.
   High speed class film/screen combination: a small decrease in perceptibility and no diffe-
   rence in the ability of giving a diagnosis.

   Preliminary conclusion: It seems that the influence of 0.1 mm added copper-filtration in
   the X-ray beam on the ability of giving a diagnose (the essence of radiodiagnostic) is
   small, except for the low speed class film/screen combinations.
   This result implies that 0.1 mm copper-filtration can be permanently installed in almost all
   the X-ray equipment.

With this study we have tried to make a balance between the dose-reduction effect of added
filtration in the X-ray beam for almost all the radiographic procedures and the negative effect
on the contrast in the image.

1. low speed class and high average gradient: AGFA Detail/AGFA MR6
   High speed class and moderate average gradient: AGFA Curix fast/AGFA HT-G

Session 2                                        11
RB Mooney and P S Thomas, Forster Green Hospital, Belfast

The Commission of the European Communities (CEC) have recommended quality criteria for
radiographic imaging of children, with the aim of defining a level of acceptability of normal
basic radiographs. The criteria were recommended in terms of image quality criteria,
reference dose levels and examples of good radiographic technique. A study was performed
to assess the value of these quality criteria when applied in a paediatric X-ray room. Initially,
a survey of patient dose and radiographic technique was carried out. Comparison of
entrance surface dose (ESD) and radiographic technique with the CEC quality criteria
highlighted a need for reduction of dose to infants and implied an increase in tube filtration.
Additional tube filtration was installed following assessment of dose reduction and image
quality with test objects, and confirmation from the paediatric radiologist that critical image
quality was not significantly altered. Tube kilovoltage was increased for infant radiographs,
and the resulting ESD fell by up to 50%. The CEC quality criteria for patient dose and
radiographic technique proved useful as a benchmark against which dose and technique in
an X-ray room can be compared, and as such can be used to help optimise patient dose.
The study highlighted the need for additional image quality criteria which could reflect
changes in clinical image quality.

Session 2                                      12
GUIDELINES FOR CLINICAL EVALUATION                                     OF      IMAGING

J V Cook, A Pettett, K Shah, S M Pablot, J Kyriou* M Fitzgerald*, Queen Mary’
Hospital for Children, Carshalton, Surrey & *St. George’ Hospital, London.

The aim of any Radiology department is to perform the most appropriate examination of the
highest quality, producing the lowest achievable dose and resulting in the correct diagnosis.

“Quality” does not only include the quality of the image produced but also that of the
request, clinical information and resulting report. All these factors need to be addressed
and in an attempt to do this a set of Paediatric Guidelines including referral criteria,
recommended radiographic technique, image criteria (based on CEC recommendations) and
achievable doses for 17 of the most common paediatric examinations have been produced
and implemented at our hospital.

Image quality assessment has included factors such as relevance of referrals, film labelling
and recording of exposure factors in addition to a graded clinical film scoring assessment of
the anatomical CEC image criteria and various physical parameters.

Following implementation, a range of dose savings has been achieved without significant
loss of quality. The Guidelines and system of image scoring will be described.

Session 2                                    13
J H Launders, R F Bury, P Hawkridge & A R Cowen, FAXiL, Leeds

Philips Thoravision is the first of a new generation of digital detectors requiring no X-ray to
visible photon conversion in the generation of images. As with any new technology it cannot
be assumed that the operating conditions which have previously been found to be the
"optimum" will be directly transferable. To determine the optimum conditions requires a
measure of the risk (radiation dose) and benefit (image quality). When selecting the
optimum conditions the dose (mAs), beam quality (kVp) and image processing have all to be
assessed. Both objective and subjective measurements of image quality have been made on
the Thoravision's performance. The objective measures of image quality are an absolute
descriptor of system performance, however, how they relate to the clinical setting has to be
assessed using subjective analysis. The results from these measurements are discussed in
relation to the effective dose received by the patient. The results suggest that the
Commission of the European Communities guidelines are not the "optimum" and more
research is required into the whole problem of optimisation. In particular, studies of much
greater scientific integrity are required before optimum techniques can be recommended to
the radiology community.

Session 2                                     14
A Rust, Regional Radiation Protection Service, Royal Surrey County Hospital,

In mammography, molybdenum has traditionally been used as anode and filter material at
28kV generating potential for all patients. This was assumed to produce a reasonable
balance between dose and image quality for the majority of breasts. However, when imaged
with the same quality radiation, very large or dense breasts are subject to higher radiation
doses than otherwise achievable and fatty breasts may be imaged with sub-optimal image
contrast. For such breasts, a different balance or relationship between dose and image
quality is appropriate. This can be partially achieved by radiographer's manual adjustment of
kV although visual assessment of breasts is unreliable.

Developments in X-ray tube design and automatic exposure control systems have led to
mammographic X-ray machines capable of tailoring the radiation quality to the individual
breast by adjustment of X-ray tube kV, filter, anode or combinations of these factors. This
allows a range of balances between dose and image quality to be achieved. The operation
of these systems will be considered and the impact upon patient dose and image contrast
discussed with reference to studies undertaken with commercial systems. Some practical
considerations for using these systems will also be presented.

Session 2                                    15
A Moore, C Clayton, Leeds General Infirmary

The Siemens Mammomat 3000 is a new generation mammography set which has two
targets (Molybdenum and Tungsten) and two filters (Molybdenum and Rhodium), and a pre-
programmed automatic factor selection mode (OPDOSE). OPDOSE aims to provide good
image quality whilst optimising the dose to the patient by suggesting one of four
programmes dependent on the breast thickness. The exposure factors (target, filter and tube
voltage) can be altered by the radiographers, however, the thickness boundaries can only
be altered by an engineer.

The mean glandular doses and image quality were assessed for a variety of factors with
different thicknesses of perspex in order to determine optimum exposure factors for the
population being imaged. Results for the image quality and mean glandular dose will be
presented for different breast thicknesses together with results from a clinical evaluation of
the mammograms. The modified factors, given in the table, have now been used clinically
for over a year.

            Siemens Default Settings              Settings Used Clinically

            Thickness (cm)     TF kV              Thickness (cm)        TF kV
            <3                 MoMo 26kV          <4                    MoMo 25kV
            3.1 - 4.5          MoMo 27kV          4.1 - 5.5             MoMo 26kV
            4.6 - 6            MoRh 27kV          5.6 - 7.5             MoRh 27kV
            >6                 WRh 26kV           >7.6                  WRh 30kV

Session 3                                    16
J H Launders & A R Cowen, FAXiL, Leeds

To be an effective tool in the screening for breast cancer mammography is required to
produce images of extremely high quality. There has been much debate in the radiological
community as to whether digital Systems, with limited spatial resolution capability, are
capable of fulfilling the requirements of the mammography screening programme. The
resolution, dynamic range and signal to noise ratio characteristics of computed radiography
systems have been measured and compared to those of conventional screen film systems.
In addition, psychophysical studies have been performed to verify the findings of the
objective measurements. The results clearly show the relationship between signal to noise
ratio and the detectability of microcalcifications. The effects of both increased and
decreased photon fluence (dose) are also demonstrated. In particular, the results show that
the large dynamic range of digital mammography systems means that the requirements of
the ALARA principle (doses should be As Low As Reasonably Achievable) are difficult to
meet. The problems associated with interpreting objective measures of image quality are
discussed in respect to the suitability and optimisation of digital radiographic systems.

Session 3                                   17

To top