BJR
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by McRobbie, D W
Right arrow Articles by Quest, R A
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by McRobbie, D W
Right arrow Articles by Quest, R A
British Journal of Radiology 75 (2002),523-531 © 2002 The British Institute of Radiology

Full Paper

Effectiveness and relevance of MR acceptance testing: results of an 8 year audit

D W McRobbie, PhD and R A Quest, MSc

Radiological Sciences Unit, The Hammersmith Hospitals NHS Trust & Imperial College, Charing Cross Hospital, Fulham Palace Road, London W6 8RF, UK


   Abstract
Top
 Abstract
Introduction
 Methods
 Results and discussion
 Conclusions
 Appendix
 References
 
The effectiveness and relevance of independent acceptance testing was assessed by means of an audit of acceptance procedures for 17 MRI systems, with field strengths in the range 0.5–1.5 T, acquired over 8 years. Signal-to-noise ratio and geometric linearity were found to be the image quality parameters most likely to fall below acceptable or expected standards. These received confirmed successful corrective action in 69% of instances. Non-uniformity, ghosting and poor fat suppression were the next most common non-compliant parameters, but yielded less satisfactory outcomes. Spatial resolution was not found to be a sensitive parameter in determining acceptability. 49% of all non-compliant parameters received verifiable corrective attention. A schedule of actual acceptance criteria is presented and shown to be reasonable. Parameter failure rates were shown not to have improved with time. A safety audit of 11 of the installations revealed the most common failings to be inadequate suite layout and poor use of signs. The mean number of safety issues per installation identified as requiring attention was 5, from a questionnaire of 100 points. A number of anecdotal errors and omissions are reported. The data support the importance of an appropriate acceptance procedure for new clinical MRI equipment and for the involvement of a suitably qualified safety adviser on the project team from the outset.


   Introduction
Top
 Abstract
 Introduction
Methods
 Results and discussion
 Conclusions
 Appendix
 References
 
In an increasingly quality driven health service, independent acceptance testing of new MR scanner installations is becoming more commonplace in both public and private sectors. The perceived benefits of such an exercise are said to include confirmation of the equipment's fitness for clinical use, verification that procurement specifications have been met and that the system is safe to use. The use of acceptance test data as a baseline for ongoing quality assurance (QA) is a further stated benefit.

MR image quality assessment, performed by medical physicists, usually forms a major part of the acceptance procedure. Although the development of MR image quality measurements has been ongoing for many years and most institutions carry out some degree of routine QA, there is, as yet, no published systematic study presenting evidence for the effectiveness of these measurements in acceptance testing. Firbank et al [1] recently presented a review of ongoing QA procedures for an individual MR system.

Protocols for image quality and QA have been published by various groups, including the European Community Concerted Action [2], the National Electrical Manufacturers Association (NEMA) [3–7], the American Association of Physicists in Medicine (AAPM) [8], the Institute of Physics and Engineering in Medicine [9] and the American College of Radiology [10]. The NEMA standards provide a range of methodologies for image quality parameter measurement including geometric distortion, uniformity and signal-to-noise ratio (SNR) [3]. In Europe, the Eurospin Test System is sometimes promoted as a standard set of test objects suitable for acceptance testing of clinical MRI equipment [11]. These test objects were designed with the specific purpose of assessing scanners for the identification and characterization of biological tissues by nuclear magnetic resonance, in a concerted research project. Their use in acceptance testing of modern MR systems is severely limited by their design, by virtue of being basically restricted to single slice mode.

There are no generally adopted published acceptance criteria, although the AAPM has published some guideline recommendations [12] and the NHS Estates Agency gave an illustration in their Health Guidance Note on MRI [13]. This paper addresses the issue of the effectiveness of an MR acceptance protocol developed at Charing Cross Hospital that uses custom designed test objects, acceptance criteria and methodologies broadly consistent with the NEMA standards. Specific questions asked were:

Other questions of interest include whether scanner performance has improved with respect to the acceptance criteria, and whether these local acceptance criteria are reasonable and merit widespread adoption.

In addition to image quality measurements, an audit was performed of installation safety aspects for 11 of the MR units. Safety issues and recommendations have been addressed by the Medical Devices Directorate (MDD) [14], the National Radiological Protection Board [15] and the International Electrotechnical Commission (IEC) [16].


   Methods
Top
 Abstract
 Introduction
 Methods
Results and discussion
 Conclusions
 Appendix
 References
 
MR equipment
17 MR systems from four manufacturers with field strengths of 0.5 T (n=2), 1.0 T (n=12) and 1.5 T (n=3) had an acceptance test within an 8 year period, as shown in Figure 1Go.



View larger version (29K):
[in this window]
[in a new window]
  		Figure 1. Number of acceptance tests performed each year on MR systems with field strengths of 0.5 T ({blacksquare}), 1.0 T () and 1.5 T ({square}).

 
Test objects
All acceptance tests utilized test objects built by the Radiation Physics workshop at Charing Cross Hospital. The geometric test object TO2A has a depth of 100 mm, allowing multislice and volumetric assessment of the main geometric image quality parameters contiguously through the phantom depth [17]. These parameters are geometric linearity and distortion, spatial resolution (line pairs per centimetre), line spread function (LSF), slice position, slice profile and width, slice angle and three-dimensional (3D) LSF (where appropriate). Figure 2Go shows a single slice through TO2A. Additionally a uniform phantom TO1A and loading ring TO0, filled with 2 g l-1 sodium chloride, permitted the comparative measurement of SNR and uniformity [18]. Ghosting was assessed using a small sample, offset in the field of view [9], and fat saturation was assessed using the ghosting bottle and a similar bottle filled with olive oil near the isocentre. All water-based test objects were filled with 7 mM nickel chloride to give a T1 and T2 of approximately 200 ms over the frequency range 21–63 MHz, with minimal temperature dependence. The loading ring was only used for SNR and uniformity measurements with object TO1A.



View larger version (134K):
[in this window]
[in a new window]
  		Figure 2. Single slice through the geometric object TO2A.

 
Image quality measurement protocols
Image quality measurements were performed according to locally derived protocols based upon extensive experience and research [19]. Definitions of measurements are given in the Appendix. These follow a hierarchical design as follows.

Parameters dependent upon coil
The SNRs for head coil, body coil, and all other coils, including phased array, were assessed for a standard spin echo (SE) sequence; TR=500 ms, TE=15 ms, 250 mm FOV, 256 x 256 matrix, and N x 5 mm slices. The number of slices N varied according to the capability of the scanner. Uniformity was assessed for head and body coils.

Parameters dependent upon sequence (for head coil)
These parameters included image SNR, geometric parameters, resolution and ghosting. Sequences evaluated included: two-dimensional (2D) multislice SE; turbo SE/fast SE (TSE/FSE); 3D (spoiled) gradient echo (SPGR); 2D turbo fast low angle shot, inversion recovery-prepped SPGR; half Fourier TSE/single shot FSE (HASTE/SS-FSE); gradient and spin echo (GRASE); and single shot echo planar imaging.

Parameters sensitive to position in magnet
These included the geometric parameters and fat saturation. They were assessed at the isocentre, and at a 200 mm offset in the coronal orientation, in a shoulder-like position. Additionally, geometric and slice parameters were assessed for the three principal image planes using SE.

In all cases acceptance tests took place after manufacturers had completed their commissioning procedures to their own satisfaction. Typically 60 image series were acquired with a total of 300–400 individual images. Image acquisition took between half a day and one whole day. An engineering representative of the manufacturer was usually present during data acquisition.

Image evaluation
Where possible images were evaluated by transferring image data by CD, Magneto-Optical Disk or network to a Sun workstation running AnalyzeTM software (Analyze Direct Inc., Lenexa, KS). Where transfer was not possible, all measurements were made on the MR console or independent workstation in situ. In these instances line profiles necessary for analysis were read out manually pixel-by-pixel from the manufacturer's line profile graphs.

Acceptance criteria
An outline of specific criteria is given in Table 1Go. These follow a two-tier scheme. Failure to meet the minimum standard constituted a hard fail and requires remediation. Failure to meet the expected value constituted a soft fail, and may require further investigation by the manufacturer, leading to re-testing or, ultimately, remediation as for a hard fail.


View this table:
[in this window]
[in a new window]
  		Table 1. Selected image quality acceptance criteria (head coil only)

 
The criteria for SNR, especially absolute SNR (ASNR), are derived from theoretical considerations for head coils [20]. For other coils the values are derived from a mixture of practical experience, manufacturer information and theory, and where none of the above existed, they are based upon the premise that a surface or specialized coil should be at least as good as the head coil.

For geometric parameters, the criteria are based upon the known linearity specifications of the gradient systems. All other criteria were based upon the authors' experience and manufacturers' published data or technical specifications.

Following analysis, a report was presented to the institution and forwarded to the manufacturer at their discretion. Follow-up testing or discussions took place at the invitation of the institution. Follow-up involving the medical physics team occurred in most, but not all, instances where criteria were not met.

Safety questionnaire
In 11 systems from 1996 onwards, a safety evaluation was completed. This was based upon a 100 point questionnaire derived from MDD and IEC publications [14, 16] and addressed the following areas.

  1. Compliance with national and international standards.
  2. Layout of suite.
  3. Magnet room features.
  4. Access to controlled area.
  5. Signs.
  6. Room safety features.
  7. Anaesthetic equipment.
  8. Security.
  9. Fire emergency features.
  10. Ancillary equipment.
  11. Operational issues: safety policy, maintenance, cryogen handling, etc.
  12. Manufacturer's documentation.
  13. Magnetic field exposure and restriction of exposure.
  14. Acoustic noise.
  15. Fringe field survey.

The questionnaire was completed with the assistance of a senior member of local MR staff, usually the superintendent, and the manufacturer where required.


   Results and discussion
Top
 Abstract
 Introduction
 Methods
 Results and discussion
Conclusions
 Appendix
 References
 
What are the most common examples of parameter non-compliance?
Figure 3Go shows the distribution of hard and soft parameter failures according to image quality parameter categories. The term "parameter failure" means that a particular criterion was not met for a defined condition, e.g. for a given coil, sequence, orientation, etc. The same system could exhibit several parameter failures, e.g. distortion or SNR etc., within an image quality category.



View larger version (19K):
[in this window]
[in a new window]
  		Figure 3. Total occurrences of hard failures ({blacksquare}) and soft failures () by image quality parameter. SNR, signal-to-noise ratio.

 
Geometric linearity/distortion and SNR produced the greatest number of hard and soft failures. Uniformity and ghosting produced an intermediate number, with the other parameters seldom failing. Spatial resolution never produced a hard failure. Only four systems had no hard failures.

Is effective corrective action taken in cases of non-compliance?
There were five categories of outcome following hard or soft failure.

  1. Replacement. A component (either hardware or software) was deemed to be faulty and required replacement. Replacement occurred in 23% of total failure occurrences.
  2. Re-calibration. Some preventative maintenance or re-installation procedure was applied, resulting in the measured parameter attaining compliance. This occurred in 11% of failure cases.
  3. Re-testing. The image quality test was repeated, perhaps with some scan parameter or physical set-up modification, and the measured parameter was compliant to the satisfaction of the tester. This occurred for 15% of failures.
  4. Unknown. The outcome was unknown to the tester. 17% of outcomes were unknown.
  5. Nil. No effective action was known to have occurred. 34% of non-compliance cases received no remediation reported back to the testers. These were mainly for soft failures but included hard failures for signal uniformity.

Grouping together categories 1–3 as favourable, a verifiable beneficial outcome was obtained for 49% of all occurrences of non-complaint measurements.

Which parameters have yielded the most effective remediation?
A breakdown of outcomes by parameter is shown in Table 2Go. Favourable outcomes were obtained for 81.8% of SNR problems, 66.7% of slice problems and 58.3% of linearity/distortion problems. In practice, SNR outcomes usually involved the replacement or retuning of a coil. Fat suppression and ghosting failures did not result in any replacement or re-calibration.


View this table:
[in this window]
[in a new window]
  		Table 2. Percentage outcomes grouped by parameter

 
Are scanners getting better?
Figure 4Go shows mean failure rates per scanner from 1994 to 2000. The one scanner investigated in 2001 had numerous problems, which were all corrected. The mean failure rates were 1.0 parameter (standard deviation (SD 0.2) for hard failures and 1.2 parameters (SD 0.5) for soft failures. All rates were within 2 SD of the mean. The very slight positive trend for all failures has a positive correlation coefficient (R) of 0.64. We conclude that results from the commissioning of scanners are not improving in terms of compliance with acceptance standards.



View larger version (21K):
[in this window]
[in a new window]
  		Figure 4. Mean failure rates per scanner from 1994 to 2000. {blacksquare}, hard failure; , soft failure.

 
Figures 5–8GoGoGoGo show measurements for head coil ASNR, linearity, distortion and slice position, ghosting and fat suppression plotted against the year of installation. There is no significant trend for improvement of SNR or slice parameters. It could be argued that geometric linearity appears to be getting worse (R=0.67), although this trend is largely caused by two outlying points. Quadrature ghosts are now generally extremely low. Fat suppression methods have improved over the study period (R=0.46).



View larger version (11K):
[in this window]
[in a new window]
  		Figure 5. Head coil absolute signal-to-noise ratio (ASNR) results by year. Higher values mean better performance. The theoretical limit for the particular phantom and sequences used are indicated by McRobbie [18]. ––, expected value; —, minimum value; – – –, theoretical maximum value.

 


View larger version (15K):
[in this window]
[in a new window]
  		Figure 6. Geometric linearity (), distortion (•) and slice position ({triangleup}) results by year. Lower values indicate better performance. A trendline for geometric linearity is shown.

 


View larger version (14K):
[in this window]
[in a new window]
  		Figure 7. Ghosting results by year. Lower values indicate better performance. The minimum for short echo time (TE) also applies to quadrature ghosts and is the expected value for long TE. {blacksquare}, short TE; {square}, long TE; {blacktriangleup}, quadrature.

 


View larger version (13K):
[in this window]
[in a new window]
  		Figure 8. Fat suppression results by year. Higher values indicate better performance. The trendline () is for fat suppression ratio. {blacksquare}, water; {blacktriangleup}, fat.

 
Are the acceptance criteria acceptable?
An audit should provide a feedback loop to facilitate quality improvement. In this section we examine numerical parameter values and assess the validity of the locally selected acceptance criteria. The ideal way to set the acceptance criteria would be to use manufacturers' data. Unfortunately this is not always available or, if it is, may only be valid for specific test methodologies. The actual criteria for minimum and expected performance were selected on the basis of experience. Only in the case of SNR can this be compared with theory. Instead we propose to show that the chosen acceptance values are reasonable. We will show that for each parameter, the expected value is close to the mean of measured values and that the minimum value broadly represents an average of the 95th (or 5th) percentile and the mean -1.7 SD (the 95% confidence value for a one-tailed distribution). The former descriptor assumes a normal distribution, the latter is more directly descriptive of the actual data, which may be non-normal as in the case of SNR. In all the figures we show results of the initial acceptance tests, prior to any remediation.

Results for head coil ASNR are shown for all 17 systems in Figure 5Go. The mean value was 12 880 Hz1/2 ml-1 T-1 (SD 2150). This is within 3% of the expected value for a measurement with an experimental error of 8% [19]. Using a one-tailed normal distribution, 95% of values should lie within 1.7 SD of the mean, or above 9 227 Hz1/2 ml-1 T-1. This is within 8% of the minimum criterion and no systems gave less ASNR than this minimum. The 5th and 95th percentile values were 10 482 Hz1/2 ml-1 T-1 and 17 893 Hz1/2 ml-1 T-1, respectively. In this study none of the systems hard failed for the head coil SNR. All SNR failures related to specialist coils and were such that diagnostic quality of clinical images would have been compromised.

It should be noted that the present use of ASNR differs from that of a previous publication [20], in that no correction is made for relaxation effects (T1 and T2) in the present paper. Applying this correction gives a mean ASNR of 13 500 Hz1/2 ml-1 T-1 with a standard deviation of 2100 Hz1/2 ml-1 T-1 for the 22 systems investigated by McRobbie [20]. The same paper quoted a theoretical maximum SNR of 23 600 Hz1/2 ml-1 T-1 for a quadrature head coil reducing to 19 900 Hz1/2 ml-1 T-1 when T1 and T2 corrections are applied for the specific sequence. This figure does not, however, include coil losses or noise in the electronics, and assumes only an inductive component and idealized radio frequency field geometry. By comparison, Redpath and Wiggins [21] reported an achievable SNR of 17 100 Hz1/2 ml-1 T-1 for one example of a transmit-receive head coil, the value dropping to 14 500 Hz1/2 ml-1 T-1 when relaxation correction is applied. These values agree sufficiently well with our proposed minimum and expected values.

An analysis is not presented here for the more specialized coils, as individual manufacturer design considerations become more important and a similar body of theoretical and published data is sparse.

Geometric values are shown in Figure 6Go. Mean values, worst values, SDs and 5th and 95th percentile values are shown in Table 3Go. The mean of means for linearity, distortion and position was 0.95% and the mean of worst values per system was 1.96%. The 95% probability values were between 1.62% and 1.95% for linearity, distortion and position, and 0.6 mm for slice width. These agree well with the minimum acceptable values of 2% and 0.5 mm, respectively. The criteria of 1%, 2% and 0.5 mm for the head coil/isocentre therefore appear to be reasonable.


View this table:
[in this window]
[in a new window]
  		Table 3. Results for geometric and slice parameters

 
Artefact-related parameter values are shown in Figure 7Go and Figure 8Go. Mean values, worst values, SD and 5th and 95th percentile values are shown in Table 4Go. The same approach as used for the geometric parameters fits well with the minimum acceptable values of 1% (short echo time (TE)) and 2% (long TE) and expected value of 1%, which reached the 95th percentile. The ghosting acceptance criteria therefore appear to be reasonable.


View this table:
[in this window]
[in a new window]
  		Table 4. Results for artefact-related parameters

 
This analysis proved more problematic for fat suppression methods where an improvement in scanner performance was evidenced by a positive correlation. The mean of 89.4% and the mean 1.7 SD calculation would suggest a much lower minimum acceptable value than that chosen (90%). However, all but four systems achieved over 90%, and only one achieved less than this since 1998. We would therefore defend the choice of 90% as being reasonable for state-of-the-art equipment. At present we have not suggested an expected value.

Data for non-suppression of the water signal by fat saturation supported the original criterion of 90% water retention. Using 1.7 SD from the mean gives a possible minimum value of 91.8%. However, the mean value was 97.4% and the modal value 100%. An inadvertent water saturation to 90% of equilibrium magnetization could have untoward consequences for the appearance of clinical images. In particular, as fat saturation is usually applied in a non-spatially selective manner, the partial water saturation may result in spurious T1 weighting. 70% of scanners achieved over 98% for this parameter and in 2000 the minimum criterion for measurement at the isocentre was changed from 90% to 98%.

What are the most common safety issues relevant to a new installation?
The mean number of failures/inadequate answers to the safety questionnaire per scanner was 4.7%. The most common failures were for lack of, or inadequate, warning signs (13%), poor layout (11%), staff/patient lockers missing or not delivered at time of testing (9%), inadequate storage (8%) and poor waveguide position (6%). Other items occurring more than once included poor coil storage, inadequate labelling of ancillary equipment, inadequate labelling of quench or stop buttons, poor visibility of the scanner from the control room, inappropriate or missing fire extinguishers, no lock on the magnet room door and the door alarm fitted but not working. Other noteworthy items that occurred only once included no emergency off button in the magnet room, intercom failure, CCTV not installed although ordered and no manufacturer's documentation. Although not safety-related, other anecdotal occurrences included the failure of a manufacturer to notice that they had not supplied any coils other than those built in, and non-installation of sequences and software that had been ordered.


   Conclusions
Top
 Abstract
 Introduction
 Methods
 Results and discussion
 Conclusions
Appendix
 References
 
In conclusion, we have demonstrated that acceptance testing is effective, leading to a successful remediation rate of at least 49%. This includes instances where coils and other hardware components were replaced. The most useful image quality parameters were SNR, geometric linearity and distortion, and the safety evaluation. Those parameters with poor remediation rates were ghosting and fat saturation. Spatial resolution was useful for characterizing k-space segmented sequences such as GRASE, HASTE and FSE/TSE, but had no bearing upon acceptance. A simple analysis of the proposed numerical acceptance criteria values has shown them to be reasonable and attainable and the criteria contained in Table 1Go are commended to the MR community.

The majority of safety issues related to design compromises in the suite layout and inadequate use of signs. The involvement of an appropriately qualified and experienced MR safety expert, the "designated professional" as proposed in the MDD guidance, at the initial planning stages is supported and recommended.


   Appendix
Top
 Abstract
 Introduction
 Methods
 Results and discussion
 Conclusions
 Appendix
References
 
Image quality parameter measurements
Signal-to-noise ratio (SNR)
The benchmark measurement was absolute SNR (ASNR) [19] measured in Hz1/2 ml-1 T-1 for the head and body coils, calculated from the mean image SNR for all slices within the specified volume (<ISNR>slices) Go


where {Delta}x, {Delta}y, {Delta}z are voxel dimensions in centimetres, NEX is number of excitations (integer) and NPE is phase encode product (integer). For two-dimensional Fourier transform, NPE is simply the number of phase encode steps. ISNR is Go


where S is the mean pixel number over a region of interest (ROI) covering 75% of the test object cross-sectional area (170 cm2) and <{sigma}> is the mean standard deviation of four artefact-free ROIs situated outside the test object. The combined ROI contained at least 400 pixels. A Raleigh distribution correction factor of 0.66 is used to express the SNR in terms of a normal distribution of noise. For non-uniform coils, a smaller signal ROI was selected commensurate with the intended clinical use of the coil.

Uniformity
Uniformity was measured from the mean integral uniformity I on a slice-to-slice basis. Go


where M is the maximum and m the minimum uniformity over an ROI encompassing 75% of the phantom area in each slice. The use of pre-measurement filtering as recommended by the National Electrical Manufacturers Association was found to be unnecessary. Additionally, a whole volume integral uniformity was measured with a volume of interest containing all slices.

Geometric linearity and distortion
Non-linearity indices Lx and Ly are defined with respect to the frequency and phase axes as Go


Go


where X is the measured distance (frequency), Y the measured distance (phase) and L the true length in the test object (180 mm), averaged over all slices.

Distortion is defined as the maximum deviation D from the mean Go


where d is the measured radial distance between points in the test object. The maximum value over the scan volume, i.e. all slices, is used (<d>slices).

Spatial resolution
Spatial resolution was assessed visually from the line pair test patterns with 2.5, 3.3 and 5 line pairs per cm, and from the measurement of modulation across the test patterns as follows: Go


Line spread functions
Line spread functions (LSF) were generated by successive subtraction of pixel values of a profile across the angled edge according to the method of Judy [22]. Full width half maxima (FWHM) were used as a measurement of spatial resolution. The asymmetric nature of the LSF precluded accurate determination of modulation transfer functions [17].

Slice position
Slice position was measured from the in-plane displacement of the shadows of the two crossed rods, and expressed as a percentage of the multislice coverage in the slice select direction.

Slice width
Slice width was measured from the rescaled geometric means of the FWHM of the slice ramp shadows.

Ghosting
Ghosting used the standard off-centre bottle technique with measurements in each quadrant of signal, phase ghost, quadrature ghost and pure background noise. The phase ghost ROI was positioned in the area of maximum artefact. Average ghost-to-signal ratios (GSRs) for all slices were calculated as Go


where g is the mean pixel value of ROI within the ghosting region, b is the mean pixel value in the ghost-free region of image and S is the mean pixel value in the primary image of the sample.


   Acknowledgments
 
The MR test objects were constructed in the Radiation Physics workshop at Charing Cross Hospital and thanks are due to Bob Reader and Ray Coote. We wish to thank the following persons, who participated in one or more of the acceptance tests: Annie Papadaki, Karin Shmueli, Mark Brewin, Alan McBride, Ruth O'Gorman, Sarah Gulliford and Jason Oates.

Received for publication July 9, 2001. Revision received November 27, 2001. Accepted for publication January 7, 2002.


   References
Top
 Abstract
 Introduction
 Methods
 Results and discussion
 Conclusions
 Appendix
 References
 

  1. Firbank MJ, Harrison RM, Williams ED, Coulthard A. Quality assurance for MRI: practical experience. Br J Radiol 2000;73:376–83.[Abstract]
  2. EEC Concerted Research Project. Identification and characterisation of biological tissues by NMR- Part IV. Protocols and test objects for the assessment of MRI equipment. Magn Reson Imaging 1988;6:195–9.[Medline]
  3. National Manufacturers Electrical Association. MS1 Determination of the signal-to-noise ratio (SNR) in diagnostic magnetic resonance images. Washington, DC: NEMA, 1988.
  4. National Manufacturers Electrical Association. MS2 Determination of two dimensional geometric distortion in diagnostic magnetic resonance images. Washington, DC: NEMA, 1990.
  5. National Manufacturers Electrical Association. MS3 Determination of image uniformity in diagnostic magnetic resonance images. Washington, DC: NEMA, 1989.
  6. National Manufacturers Electrical Association. MS5 Determination of slice thickness in diagnostic magnetic resonance images. Washington, DC: NEMA, 1992.
  7. National Manufacturers Electrical Association. MS6 Characterization of special purpose coils for diagnostic magnetic resonance images. Washington, DC: NEMA, 1991.
  8. Price RR, Axel L, Morgan T, Newman R, Perman W, Schneiders N, et al. Quality assurance methods and phantoms for magnetic resonance imaging: Report of AAPM Nuclear Magnetic Resonance Task Group No 1a. Med Phys 1990;17:287–95.[Medline]
  9. Lerski RA, de Wilde J, Boyce D, Ridgeway J. Quality control in magnetic resonance imaging, IPEM Report 80. York: The Institute of Physics and Engineering in Medicine, 1998.
  10. American College of Radiology. ACR MRI Quality Control Manual. Reston, VA: The American College of Radiology, 2001.
  11. Lerski RA, McRobbie DW, Straughan K, Walker PM, de Certainnes J, Bernard AM. Identification and characterisation of biological tissues by NMR- Part V. Multi-centre trial with protocols and prototype test objects for the assessment of MRI equipment. Magn Reson Imaging 1988;6:201–14.[Medline]
  12. Och JG, Clarke GD, Sobol WT, Rosen CW, Mun SK. Acceptance testing of magnetic resonance imaging systems: report of AAPM Nuclear Magnetic Resonance Task Group No 6a. Med Phys 1992;19:217–29.[Medline]
  13. NHS Estates Agency. HGN Magnetic Resonance Imaging Health Guidance Note. London: HMSO, 1997.
  14. Medical Devices Directorate. Guidelines for magnetic resonance diagnostic equipment in clinical use with particular reference to safety. London: HMSO, 1993.
  15. National Radiological Protection Board. Limits on patient exposure during clinical magnetic resonance procedures. Documents of the NRPB. Chilton, UK: NRPB, 1991:2(1).
  16. International Electrotechnical Commission. IEC 60601-2-33 Medical Electrical Equipment—Part 2. Particular requirements for the safety of magnetic resonance equipment for medical diagnosis. Geneva, Switzerland: IEC, 1996.
  17. McRobbie DW. A 3-D volumetric test object for geometry evaluation in MRI. Med Phys 1997;24:743–50.[Medline]
  18. McRobbie DW. Radio frequency eddy current losses for an annular conductor: theory and applications. Med Phys 1993;20:1555–60.[Medline]
  19. McRobbie DW. Quantitative assessment of magnetic resonance imaging systems. PhD Thesis. London: University of London, 1992.
  20. McRobbie DW. The absolute signal-to-noise ratio in MRI acceptance testing. Br J Radiol 1996;69:1045–8.[Abstract]
  21. Redpath TW, Wiggins CJ. Estimating achievable signal-to-noise ratios of MRI transmit-receive coils from radiofrequency power measurements: applications in quality control. Phys Med Biol 2000;45:217–27.[Medline]
  22. Judy PF. The line spread function and modulation transfer function of a computed tomography scanner. Med Phys 1976;3:233–6.[Medline]



This article has been cited by other articles:


Home page
 Br. J. Radiol.Home page
C J Koller, J P Eatough, P J Mountford, and G Frain
A survey of MRI quality assurance programmes.
Br. J. Radiol., July 1, 2006; 79(943): 592 - 596.
[Abstract] [Full Text] [PDF]

Home page
 RadioGraphicsHome page
J. Zhuo and R. P. Gullapalli
AAPM/RSNA Physics Tutorial for Residents: MR Artifacts, Safety, and Quality Control
RadioGraphics, January 1, 2006; 26(1): 275 - 297.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by McRobbie, D W
Right arrow Articles by Quest, R A
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by McRobbie, D W
Right arrow Articles by Quest, R A

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
BJR DMFR IMAGING  ALL BIR JOURNALS