A multi-center inter-manufacturer study of the temporal stability of phase-contrast velocity mapping background offset errors
© Gatehouse et al.; licensee BioMed Central Ltd. 2012
Received: 14 October 2011
Accepted: 1 October 2012
Published: 20 October 2012
Phase-contrast velocity images often contain a background or baseline offset error, which adds an unknown offset to the measured velocities. For accurate flow measurements, this offset must be shown negligible or corrected. Some correction techniques depend on replicating the clinical flow acquisition using a uniform stationary phantom, in order to measure the baseline offset at the region of interest and subtract it from the clinical study. Such techniques assume that the background offset is stable over the time of a patient scan, or even longer if the phantom scans are acquired later, or derived from pre-stored background correction images. There is no published evidence regarding temporal stability of the background offset.
This study assessed the temporal stability of the background offset on 3 different manufacturers’ scanners over 8 weeks, using a retrospectively-gated phase-contrast cine acquisition with fixed parameters and at a fixed location, repeated 5 times in rapid succession each week. A significant offset was defined as 0.6 cm/s within 50 mm of isocenter, based upon an accuracy of 10% in a typical cardiac shunt measurement.
Over the 5 repeated cine acquisitions, temporal drift in the baseline offset was insignificant on two machines (0.3 cm/s, 0.2 cm/s), and marginally insignificant on the third machine (0.5 cm/s) due to an apparent heating effect. Over a longer timescale of 8 weeks, insignificant drift (0.4 cm/s) occurred on one, with larger drifts (0.9 cm/s, 0.6 cm/s) on the other machines.
During a typical patient study, background drift was insignificant. Extended high gradient power scanning with work requires care to avoid drift on some machines. Over the longer term of 8 weeks, significant drift is likely, preventing accurate correction by delayed phantom corrections or derivation from pre-stored background offset data.
Phase-contrast velocity mapping is applied to blood flow and myocardial velocity mapping in cardiovascular magnetic resonance (CMR). Velocity images are formed by subtracting the phase images of two acquisitions with differing velocity sensitivity. As is well known, other phase differences between the two images cause stationary tissue to display an apparent non-zero velocity, known as the background offset or baseline error. This may vary gradually with position over the velocity image and underlies stationary and moving tissues. Maxwell (or concomitant) gradient effects are one cause which can be corrected analytically. Eddy currents are another cause, and these are corrected to a large extent by actively-shielded gradient coils and pre-emphasis[4, 5] although this is complicated by mechanical vibration effects. The remaining cause of background offset is residual errors in the pre-emphasis which accumulate phase errors during the time between velocity-encoding and echo. The background offset can sometimes be reduced by running some functions of the pulse sequence more slowly, but this is often incompatible with applications such as breath-hold imaging.
The background offset error may be corrected using the apparent velocity values in stationary tissue close to the vessel of interest. This correction is sufficient in most applications outside the thorax. However, for flow quantification near the heart adjacent stationary tissue is often absent. This can be solved by determining the offset in stationary tissue from the whole FOV and subtracting it using a fitted surface[7, 8], thereby correcting the offset at the location of the vessel of interest. This approach is limited by several requirements: sufficient stationary tissue at adequate SNR, avoidance of phase-encode FOV wraparound, avoiding steady venous flow in the stationary tissue mask, and elimination of signal far from isocenter where the error may be highly nonlinear. All of these factors hinder reliable automatic correction by this approach.
A more basic method measures the background offset by repeating the flow study using a uniform stationary phantom after the patient, with identical sequence parameters, and is currently considered the “gold standard” for correction. The phantom measurements are subtracted from the patient’s velocity study to correct the background offset. This approach is simple but it is inconvenient at clinical centres and might be circumvented by a pre-stored set of corrections for a sufficiently broad range of plane orientations and image parameters. Evidently, both correction methods demand temporal stability of the velocity offset.
This work evaluated the temporal stability of phase-contrast background velocity offsets where there appear to be no previous publications.
The temporal stability of the background error might be questioned on three timescales. First, is it stable over a long period enabling the use of pre-stored corrections? Second, is the background error stable during the patient session, so that a phantom scan acquired at some time after the patient session is a reliable measurement of the error during a clinical flow study? Third, is the background error stable during all the images of a cine velocity map, even where it was retrospectively gated with continuous gradient waveform activity? The third question was confirmed long ago. In the absence of published work answering the first two questions, we aimed to measure the temporal stability of background offset errors.
CMR test phantoms
The specification was to measure apparent velocity in a stationary uniform phantom, up to 50 mm distance from isocenter as measured in the image plane. Each centre used a different test phantom of gadolinium doped gelatine, or water which was allowed to settle for 5 minutes before beginning scanning. In one instance, to confirm the fluid’s stability, repeated phase-contrast scans were acquired during the settling time, showing that fluid settled to < 0.1 cm/s.
CMR systems and phase contrast velocity acquisitions
The study was limited to 1.5T whole body systems as this is currently the most widely-used main field strength for cardiac CMR. Pre-emphasis for reduction of eddy current effects and automatic correction of concomitant gradient terms were employed as implemented by the vendor, whereas any other filtering or correction of background offset errors was turned off because this would perform unrealistically well on large uniform phantoms. Three 1.5T scanners were used, one from each of three manufacturers (GE Signa Excite, Philips Intera and Siemens Avanto) placed in random order as scanner types 1, 2 and 3. All systems had water-cooled gradient coils and gradient amplifiers. Two of them had helium recirculation for the cryogenic magnet, while the third had no helium fills during this work (see Discussion). Each centre acquired the same single plane at 45° oblique between transverse and sagittal planes, with anterior-posterior phase-encoding, at isocentre (the aortic plane shown in reference). This slice orientation is used clinically, and showed in an earlier study to be sensitive to offsets. On each machine, the same plane was repeated 5 times in rapid succession during each approximately weekly session, over a period of 8 weeks. No CMR service engineering recalibrations were performed during this time. Before beginning each weekly session, a delay of at least 10 minutes without scanning was allowed. Short “scout” images were acquired each week before the first phase-contrast acquisition.
On six machines, two from each of the 3 manufacturers, a further study examined the effect of a high gradient-power sequence run continuously for at least 5 minutes between two sets of 5 cine studies as acquired above. This was acquired at a later date than the main 8-week data, following initial evaluation of that data, and two machines of each type were used to test the consistency of their behaviour.
Pulse sequence parameters
Scanner type 1
Scanner type 2
Scanner type 3
Slice thickness [mm]
TR, TE [ms]
Pixel size (acq) FExPE [mm]
1.25 x 2.5
1.2 x 2.4
1.25 x 2.5
FE pixels (excl oversampling)
PE pixels (acq)
Raw data [lines/cycle]
Flip angle [deg]
GE Signa Excite at VUMC Amsterdam NL, and Addenbrooke’s Hospital Cambridge, UK. The readout ADC bandwidth was 41.67 kHz (pixel bandwidth 326 Hz/pixel). This used symmetric velocity-encoding (i.e. two sequence repetitions with positive and negative velocity sensitivities around the velocity-compensated waveform, also known as “balanced” velocity-encoding). The “flow analysis” flag was on, disabling a spatial high-pass filter used for phase-contrast angiography background suppression. The GE “flow optimization” control resulted in longer TE and TR than the other scanners and was therefore not used. Twenty cardiac phases were reconstructed in a 600 ms simulated R-R interval (i.e. temporal interpolation was applied by reconstruction).
Philips Intera. Both systems were at Lund University Hospital, Sweden. Pixel bandwidth 355 Hz/pixel (Fat/Water Shift 0.62 pixels). The “Gradient mode” was set to “Default” and “PNS mode” to “Moderate”. The slice-selective RF pulse used an asymmetric design with a late centre. The background phase-offset correction (“LPC filter”) was obviously switched off for this study. Fifteen cardiac phases were reconstructed (i.e. temporal interpolation was not performed during reconstruction).
Siemens Avanto. Both systems were at Royal Brompton Hospital, London, UK. The controls for RF pulse and gradient mode, which control the use of faster and stronger RF and gradient pulses, were both set to “Normal” mode. This sequence was run ungated to enable a rapid weekly test of background errors in multiple image planes at multiple slice shifts from isocenter, i.e. it did not produce a cine for each slice. The images from all other planes except the transverse-to-sagittal slice were subsequently left unused, but the total session time for the 5 repeated scans was similar to that of the other scanners. For assessment of the temporal variability during a cine sequence, a cine version of the above sequence was run once. The cine reconstruction interpolated 20 cine frames into the 1 second simulated R-R interval.
Two timescales of variations in the offsets were considered: among the 8 weekly sessions, and during each session of five scans. For these purposes, the mean over all the frames of cine acquisitions was taken for each ROI and the standard-deviation over these frames was also recorded. The standard-deviation over the cine frames was used only to check whether the background offset variations exceeded the level of random noise and systematic errors in the retro-gated cine reconstruction.
The following procedure was adopted for the purpose of determining a significant background drift. The largest peak-to-peak differences in the offsets were measured within each group of 5 scans and also across all 40 scans. This gave the largest error which could have occurred if the stationary phantom correction method had been used at any time during the weekly scans or the full 8 weeks of data collection. The errors were compared against the 0.6 cm/s offset which was suggested as liable to affect some parameters derived from clinical flow measurements.
Variability results of background offsets
1 2 3 4
1 2 3 4
1 2 3 4
Long-term (stdev cm/s)
0.07 0.07 0.07 0.08
0.18 0.18 0.17 0.21
0.16 0.08 0.07 0.12
Long-term (pk-pk cm/s)
0.35 0.29 0.28 0.37
0.64 0.62 0.60 0.86
0.61 0.34 0.30 0.48
Session (pk-pk cm/s)
0.25 0.16 0.28 0.20
0.47 0.48 0.40 0.50
0.19 0.21 0.19 0.23
Cine (Sc) (stdev cm/s)
0.12 0.13 0.12 0.12
0.06 0.05 0.05 0.05
0.06 0.08 0.04 0.07
Spatial (Ss) (stdev cm/s)
0.94 1.0 1.1 1.2
0.20 0.18 0.32 0.21
0.23 0.32 0.22 0.28
Recalling the aim of reliable background offset correction for routine clinical use of phase-contrast flow measurements, we reconsider the correction methods explained in the introduction. The idea of pre-stored corrections appeared unlikely to be reliable on 2 of the 3 scanners tested, on which the peak-peak variation over 8 weeks approached or exceeded the 0.6 cm/s regarded as liable to introduce clinically significant errors in some situations. The absence of any obvious drift during the 8 weeks implies that any changes in helium levels over this time had negligible effect. Over a shorter timescale, for correction using a stationary phantom after the patient, this was reliable to well below the 0.6 cm/s limit in all scanner types, although the drift observed in scanner type 2 might eventually violate this limit.
There was no significant impact of the high power scan for scanner types 1 and 3, where the difference was <0.6cm/s. Also an earlier inter-scanner study of gradient amplifier response delays has shown high stability. However, for scanner type 2, a larger change occurred. The drift effect in scanner type 2 was perhaps related to heating of system components such as the gradient amplifiers, or alterations in the residual eddy-currents, as was shown by the effect of the high power scans. This type of drift might be expected until gradient-related heating reaches a temperature steady-state against the cooling system. For the interleaved acquisition of the two sets of phase images used in the phase-contrast technique, the effects of main field-distortion known to arise from heating of the passive shim steel should almost exactly cancel. However, small sequence timing differences, even including programming errors, between the two sets of phase images may exacerbate sensitivity. Furthermore, it can be shown that gradient anisotropy (i.e. of inter-axis delays) has no impact on background offset errors even for oblique planes. Probably the main factor is the temporal stability of gradient amplifier performance, and also the stability of eddy-current generation and their compensation.
The background offset error may reach clinical significance only in certain situations. The most typical of these is where the through-plane velocity data is integrated over a vessel cross-section, then summed again over all the cine frames the cardiac cycle to obtain cardiac flow measurements. The clinical impact of this error may increase with the large cross-sectional area of dilated vessels, for patients with a slow heart-rate (long R-R interval) and also for measurements derived from multiple flow acquisitions, such as the Qp/Qs ratio. The maximum allowable offset of 0.6 cm/s, as considered in this study, is related to these measurements. It relates to a typical stroke volume error of less than 5%, or in terms of a Qp/Qs ratio error below 10%. However, in other applications such as peak velocity assessment the background offset is usually negligible.
This study implies that correction using the clinical phase-contrast scan itself is preferable, since this would annul the question of drift during the time between clinical application and a separate correction scan, but this demands further work on the automated background correction or field-camera methods. The results also suggest cautious use of phantom corrections as a reference for evaluating the performance of such methods.
For this work the relevant value for validation of robustness against noise was the variation in the mean ROI value when making repeated measurements. The spatial standard deviation (Ss) shown in Table2 was larger and is explained as follows. The theoretical standard deviation of repeated measurements of the ROI mean (Sm) is given by the spatial standard deviation Ss divided by the square root of the number of independent pixels (Npix) in the ROI, where Npix was 226 (245 on type 2). This assumes any slope in background error over the ROI is << Ss, and also that noise is spatially and temporally uncorrelated. The value Sm gives the random noise lower limit on the standard-deviation during the cine (Sc), because Sc also includes systematic variations between frames, for example transient effects after the phase-encoding is advanced. Using the mean of Ss over the four ROIs, the values of Sm for types 1,2,3 were 0.07cm/s, 0.02cm/s, 0.02cm/s; these were substantially below the level of significance and support the robustness of this work against noise.
This study was limited to 1.5T systems as most CMR is acquired at this field strength, and there was no reason to expect a major difference in stability at other field strengths using superconducting magnets. The study was also limited to phantom acquisitions, and a more “real-life” study of changes in these after patient scanning might be more relevant. However, phantom acquisitions were essential to establish the underlying stability of the scanners. In vivo, correction algorithms are often applied. However, with the phantoms used in this study the evaluation of such algorithms is not realistic for in-vivo situations, as the most critical issue is not simulated well i.e. the lack of stationary tissue around the vessels of interest.
Over the duration and activity of a typical patient study, background drift was insignificant. However, the combination of extended high gradient power scanning and work requiring background correction requires care to avoid drift on some machines. Thus, on most systems, the background offset in the velocity measurement can be corrected by a separate phantom scan within the same imaging session. Over the longer term of 8 weeks, significant drift in offset value is likely, preventing accurate correction by delayed phantom correction scans or derivation from pre-stored background offset images.
KMB acknowledges the generous support of the Swedish Foundation for Strategic Research. PG acknowledges the support of the National Institute for Health Research Cardiovascular Biomedical Research Unit of the Royal Brompton Hospital London UK.
- Srichai MB, Lim RP, Wong S, Lee VS: Cardiovascular applications of phase-contrast MRI. AJR Am J Roentgenol. 2009, 192 (3): 662-75. 10.2214/AJR.07.3744. ReviewView ArticlePubMedGoogle Scholar
- Jung B, Föll D, Böttler P, Petersen S, Hennig J, Markl M: Detailed analysis of myocardial motion in volunteers and patients using high-temporal-resolution MR tissue phase mapping. J Magn Reson Imaging. 2006, 24 (5): 1033-9. 10.1002/jmri.20703.View ArticlePubMedGoogle Scholar
- Bernstein MA, Zhou XJ, Polzin JA, King KF, Ganin A, Pelc NJ, Glover GH: Concomitant gradient terms in phase contrast MR: analysis and correction. Magn Reson Med. 1998, 39 (2): 300-8. 10.1002/mrm.1910390218.View ArticlePubMedGoogle Scholar
- Jehenson P, Westphal M, Schuff N: Analytical method for the compensation of eddy-current effects induced by pulsed magnetic field gradients in NMR systems. JMR. 1990, 90: 264-278.Google Scholar
- Van Vaals J, Bergman A: Optimization of eddy-current compensation. JMR. 1990, 90: 52-70.Google Scholar
- Rolf MP, Hofman MB, Gatehouse PD, Markenroth-Bloch K, Heymans MW, Ebbers T, Graves MJ, Totman JJ, Werner B, van Rossum AC, Kilner PJ, Heethaar RM: Sequence optimization to reduce velocity offsets in cardiovascular magnetic resonance volume flow quantification - A multi-vendor study. J Cardiovasc Magn Reson. 2011, 13: 18-10.1186/1532-429X-13-18.PubMed CentralView ArticlePubMedGoogle Scholar
- Walker PG, Cranney GB, Scheidegger MB, Waseleski G, Pohost GM, Yoganathan AP: Semiautomated method for noise reduction and background phase error correction in MR phase velocity data. J Magn Reson Imaging. 1993, 3: 521-530. 10.1002/jmri.1880030315.View ArticlePubMedGoogle Scholar
- Lankhaar JW, Hofman MB, Marcus JT, Zwanenburg JJ, Faes TJ, Vonk-Noordegraaf A: Correction of phase offset errors in main pulmonary artery flow quantification. J Magn Reson Imaging. 2005, 22 (1): 73-9. 10.1002/jmri.20361.View ArticlePubMedGoogle Scholar
- Chernobelsky A, Shubayev O, Comeau CR, Wolff SD: Baseline correction of phase contrast images improves quantification of blood flow in the great vessels. J Cardiovasc Magn Reson. 2007, 9 (4): 681-5. 10.1080/10976640601187588.View ArticlePubMedGoogle Scholar
- Søndergaard L, Ståhlberg F, Thomsen C, Spraggins TA, Gymoese E, Malmgren L, Müller E, Henriksen O: Comparison between retrospective gating and ECG triggering in magnetic resonance velocity mapping. Magn Reson Imaging. 1993, 11 (4): 533-7. 10.1016/0730-725X(93)90472-P.View ArticlePubMedGoogle Scholar
- Gatehouse PD, Rolf MP, Graves MJ, Hofman MB, Totman J, Werner B, Quest RA, Liu Y, von Spiczak J, Dieringer M, Firmin DN, van Rossum A, Lombardi M, Schwitter J, Schulz-Menger J, Kilner PJ: Flow measurement by cardiovascular magnetic resonance: a multi-centre multi-vendor study of background phase offset errors that can compromise the accuracy of derived regurgitant or shunt flow measurements. J Cardiovasc Magn Reson. 2010, 12: 5-10.1186/1532-429X-12-5.PubMed CentralView ArticlePubMedGoogle Scholar
- Mao SS, Ahmadi N, Shah B, Beckmann D, Chen A, Ngo L, Flores FR, Gao YL, Budoff MJ: Normal thoracic aorta diameter on cardiac computed tomography in healthy asymptomatic adults: impact of age and gender. Acad Radiol. 2008, 15: 827-34. 10.1016/j.acra.2008.02.001.PubMed CentralView ArticlePubMedGoogle Scholar
- Brodsky EK, Samsonov AA, Block WF: Characterizing and correcting gradient errors in non-cartesian imaging: Are gradient errors linear and time-invariant (LTI)?. Magn Reson Med. 2009, 62 (6): 1466-76. 10.1002/mrm.22100.PubMed CentralView ArticlePubMedGoogle Scholar
- Aldefeld B, Börnert P: Effects of gradient anisotropy in MRI. Magn Reson Med. 1998, 39 (4): 606-14. 10.1002/mrm.1910390414.View ArticlePubMedGoogle Scholar
- Giese D, Haeberlin M, Barmet C, Schaeffter T, Pruessmann KP, Kozerke S: Analysis and correction of background velocity offsets in cine phase-contrast imaging using magnetic field monitoring. Magn Reson Med. 2012, 67 (5): 1294-302. 10.1002/mrm.23111.View ArticlePubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.