Introduction
Materials and methods
Description of the survey process
Construction of the questionnaire
Panel selection
Interpretation of the survey results
Results
Results of survey
BOLD MRI | |
---|---|
Preparation | Normal hydration (100 ml water), 4 h fasting from food |
Field strength | 1.5 T or 3.0 T, 3Tpreferred if available |
Sequence | 2D mGRE |
Orientation | Coronal oblique to kidneys |
In-plane resolution | 2–3 mm |
Slice thickness | 3–5 mm |
Coverage | 3–5 slices centered on renal hilum |
Parallel imaging factor | 2 |
Fat suppression | Yes |
TR (s) | 60–75 ms |
TE (ms) | 8–16 echoes, up to 50 ms (~T2* cortex) at 3T with choice of in phase for fat-water |
Averages | 1 |
Breathing mode | Breath hold |
Image quality control | Recommended |
ROI placement | Manual |
Cortical ROI | 1 stripe/slice; > 3 slices |
Medullary ROI | 3 samples/slice; > 3 slices |
Fitting | Monoexponential or log-linear |
Reporting | Cortex and medulla |
Reported metric | R2* (s−1) |
Metric statistics reporting | Mean, median, standard deviation, ROI size |
Map format | Color or grayscale quantitative map |
Respondents for Round 1 | n = 10 | Percentage |
---|---|---|
Physicists | 7 | 70 |
Clinicians | 3 | 30 |
Radiologists | 1 | 10 |
Nephrologists | 2 | 20 |
Physiologistsa | 1 | 10 |
Respondents for Round 2 | n = 24 | Percentage |
---|---|---|
Physicists | 12 (50%) | 50 |
Clinicians/physiologists | 13 (50%) | 54 |
Radiologistsa | 11 | 46 |
Nephrologists | 2 | 8 |
Physiologistsa | 1 | 4 |
Consensus/tendency | Overall (%) | Physicists (%) | Clinicians (%) | |||||||
---|---|---|---|---|---|---|---|---|---|---|
Agree | Disagree | Abstain | Agree | Disagree | Abstain | Agree | Disagree | Abstain | ||
1. Subjects’ diet needs to be controlled before the scan | Agree | 71 | 29 | 13 | 60 | 40 | 17 | 82 | 18 | 15 |
2. Subject should be scanned in a normal hydration status when clinically appropriate | AGREE | 87 | 13 | 4 | 91 | 9 | 8 | 85 | 15 | 0 |
3. Subjects are required to follow a controlled and standardized salt intake before the scan | Disagree | 35 | 65 | 29 | 25 | 75 | 33 | 44 | 56 | 31 |
Consensus/tendency | Overall (%) | Physicists (%) | Clinicians | |||||||
---|---|---|---|---|---|---|---|---|---|---|
Agree | Disagree | Abstain | Agree | Disagree | Abstain | Agree | Disagree | Abstain | ||
4. 3T is better than 1.5T for acquiring quality BOLD data | Agree | 74 | 26 | 21 | 75 | 25 | 33 | 73 | 27 | 15 |
5. The longest TE should be equal to the largest T2* expected in the kidney | Unresolved | 53 | 47 | 38 | 55 | 45 | 8 | 40 | 60 | 62 |
6. The largest TE should be equal to 1.5 times the longest T2* expected in the kidney | Unresolved | 53 | 47 | 38 | 55 | 45 | 8 | 60 | 40 | 62 |
7. Is the abdominal standard shim performed automatically by your scanner sufficient for quality BOLD data? | Disagree | 35 | 65 | 17 | 36 | 64 | 8 | 30 | 70 | 23 |
8. For quality BOLD data it is necessary to shim on a restricted volume (a “box” selected by the operator) that includes both kidneys | AGREE | 76 | 24 | 13 | 64 | 36 | 8 | 91 | 9 | 15 |
9. For quality BOLD data it is necessary to shim on a restricted volume (a “box” selected by the operator) that includes only 1 kidney (accepting a poorer shim on the other kidney) | DISAGREE | 21 | 79 | 17 | 25 | 75 | 0 | 11 | 78 | 31 |
10. Acquiring a 3D B0 map is important and useful for quality control of BOLD data | AGREE | 80 | 20 | 38 | 86 | 14 | 42 | 67 | 33 | 31 |
11. Reducing the effect of intra-renal fat improves BOLD data quality | Agree | 69 | 31 | 33 | 78 | 22 | 25 | 50 | 50 | 38 |
12. For reducing the effect of fat, choosing fat and water to be in-phase for all echoes is preferable to using fat saturation | Agree | 71 | 29 | 42 | 78 | 22 | 25 | 50 | 50 | 54 |
Most current reports on renal BOLD MRI have used 2 × 2 × 5 mm3 voxel size. There is evidence to suggest spatial resolution (i.e. voxel size) influences R2* estimates (PMID: 23571833). Please consider the evidence in this publication when responding to the following 4 questions, and keep your answers consistent | ||||||||||
13. Should BOLD acquisitions strive for a nearly isotropic resolution (e.g to help reduce unwanted effects of macroscopic field inhomogeneities)? | Agree | 67 | 33 | 25 | 64 | 36 | 8 | 63 | 38 | 38 |
14. If you agree resolution should be isotropic, do you agree that 3 × 3× 3 mm3 is a sufficient voxel size to balance SNR and resolution? | AGREE | 79 | 21 | 36 | 86 | 14 | 30 | 71 | 29 | 46 |
15. Considering your answer to the previous question, 2 × 2 × 5 mm3 is a sufficient voxel size for renal BOLD to balance SNR and resolution | Agree | 65 | 35 | 29 | 56 | 44 | 25 | 75 | 25 | 38 |
16. The right balance between parallel acceleration factor and SNR is a 2x acceleration factor for BOLD MRI | AGREE | 93 | 7 | 38 | 100 | 0 | 33 | 86 | 14 | 46 |
Consensus/tendency | Overall (%) | Physicists (%) | Clinicians | |||||||
---|---|---|---|---|---|---|---|---|---|---|
Agree | Disagree | Abstain | Agree | Disagree | Abstain | Agree | Disagree | Abstain | ||
17. The marker of interest from renal BOLD is R2* | AGREE | 100 | 0 | 0 | 100 | 0 | 0 | 100 | 0 | 0 |
18. Both exponential and log-linear fitting of the BOLD signal are acceptable for extracting R2* values | AGREE | 83 | 17 | 4 | 75 | 25 | 0 | 83 | 17 | 8 |
19. Manual ROIs drawn on BOLD with anatomical T2or T1weighted images as reference, in collaboration with a radiologist, is a suitable analysis method for a novice user | AGREE | 87 | 13 | 4 | 82 | 18 | 8 | 92 | 8 | 8 |
20. Semi-automated segmentation of cortex and medulla (e.g. PMID: 28959212, using histogram analysis to define masks for cortex and medulla based on T1maps) is a preferred analysis method | Agree | 72 | 28 | 25 | 70 | 30 | 17 | 75 | 25 | 38 |
21. Semi-automated analysis which identified pixels with “hypoxia” based on a predetermined threshold R2* (PMID: 23788716) value within the kidney is a preferred analysis method | DISAGREE | 25 | 75 | 33 | 0 | 100 | 33 | 50 | 50 | 38 |
22. Histogram analysis of T 2*/R2* maps is a preferred analysis method | Unresolved | 50 | 50 | 42 | 25 | 75 | 33 | 83 | 17 | 54 |
23. The TLCO or onion peel (PMID: 27798200) is a preferred analysis method | Unresolved | 53 | 47 | 38 | 71 | 29 | 42 | 38 | 63 | 38 |
Consensus/tendency | Overall (%) | Physicists (%) | Clinicians (%) | |||||||
---|---|---|---|---|---|---|---|---|---|---|
Agree | Disagree | Abstain | Agree | Disagree | Abstain | Agree | Disagree | Abstain | ||
24. Given the consensus that R2* is a relative measure,R2* measures oxygen availability rather than absolute oxygenation (e.g. partial pressure of oxygen) in the kidney | AGREE | 87 | 13 | 4 | 82 | 18 | 8 | 85 | 15 | 0 |
25. Renal BOLD MRI is potentially a diagnostic technique | AGREE | 78 | 22 | 4 | 64 | 36 | 8 | 92 | 8 | 0 |
26. Renal BOLD MRI is currently only useful as a research tool | Agree | 74 | 26 | 4 | 75 | 25 | 0 | 75 | 25 | 8 |
27. The most proven application for renal BOLD MRI is to study acute responses to physiological or pharmacological maneuvers | AGREE | 86 | 14 | 13 | 100 | 0 | 17 | 75 | 25 | 8 |
28. From published data to-date, renal BOLD MRI can be used to evaluate acute kidney dysfunction (acute kidney injury, allograft acute rejection, allograft acute tubular necrosis) | Agree | 67 | 33 | 25 | 75 | 25 | 33 | 60 | 40 | 23 |
29. From published data to-date, renal BOLD MRI can be used to evaluate chronic kidney dysfunction (interstitial fibrosis and tubular atrophy) | Unresolved | 53 | 47 | 21 | 44 | 56 | 25 | 60 | 40 | 23 |
30. From published data to-date, has renal BOLD MRI been shown to have sufficient (short and long term) scan-re-scan repeatability for translational research studies? | AGREE | 85 | 15 | 17 | 100 | 0 | 17 | 70 | 30 | 23 |
31. From published data to-date, has renal BOLD MRI been shown to have sufficient (short and long term) scan-re-scan repeatability for multi-site clinical trials? | Disagree | 32 | 68 | 21 | 44 | 56 | 25 | 18 | 82 | 15 |
Overall (%) | |
---|---|
1.5T is acceptable for obtaining quality BOLD data. [choose a single option] | |
Agree (comment in other) | 100 |
Disagree (comment in other) | 0 |
Do not know | 30 |
Other | 0 |
The smallest bandwidth should be chosen for the selected maximum TE | |
Agree (comment in other) | 83 |
Disagree (comment in other) | 16 |
Do not know | 30 |
Other | 10 |
Do number of echoes and/or spacing between TE’s have an effect on data quality? | |
Agree (comment in other) | 88 |
Disagree (comment in other) | 12 |
Do not know | 20 |
Other | 12.5 |
Is breath-holding the best approach for controlling motion in renal BOLD MRI? | |
Yes | 80 |
No | 20 |
Do not know | 0 |
Other | 0 |
Is analysis of cortex and medulla important? | |
Yes, both equally important | 90 |
Cortex is more important than medulla (comment in other) | 0 |
Medulla is more important than cortex (comment in other) | 10 |
Do not know | 0 |
Other | 0 |
Is renal BOLD MRI potentially a prognostic technique? | |
Yes | 100 |
No | 0 |
Do not know | 20 |
Other | 0 |
Does renal BOLD MRI reflect intra-renal oxygenation (qualitatively or quantitatively)? | |
Yes, BOLD MRI can quantify intrarenal oxygenation | 20 |
Yes, qualitatively, there are too many confounding factors for it to be quantitative | 80 |
No, it does not | 0 |
Do not know | 0 |
Other | 0 |
From existing data, do you believe renal BOLD MRI is more suited to detect changes in renal oxygenation within the same kidney rather than comparing cohorts? | |
Yes, best used to detect changes within the same kidney, due to confounding effects of blood volume and hematocrit that are different between patients | 88 |
It is suitable to compare cohorts without adjustments for patients’ blood volume and hematocrit | 0 |
It is suitable to compare cohorts with adjustment for patients’ blood volume and hematocrit | 11 |
Do not know | 10 |
Other | 0 |
Does the furosemide stress test/ physiological challenge have value? | |
Yes, always | 25 |
Yes, but it may not be necessary or suitable for some applications (comment in Other) | 75 |
No (comment in Other) | 0 |
Do not know | 20 |
Other | 0 |
Final recommendations
-
Regarding patient preparation, irrespective of the level of consensus reached, users should consider ways of normalizing the baseline physiological status. Even though data exists only for hydration status, food intake commonly is associated with fluid intake and a likely confounding factor. When feasible, fasting for 4 h may be a good option. A related issue of bowel gas has not yet been fully resolved in terms of minimizing its presence.
-
Even though the choice of 3T did not reach the 75% threshold, there is a general acceptance that, if given a choice between the field strengths, 3T is preferable. In the absence of access to 3T, 1.5T remains an adequate choice. While there is a concern for increased magnetic field inhomogeneities at 3T, there is considerable experience to-date that supports the higher sensitivity and signal-to-noise ratio (SNR) afforded by 3T, making it a preferred choice.
-
Magnetic field inhomogeneities (e.g. due to poor shimming, presence of metal or air interfaces) play an important role and should be minimized. These will affect the measured R2* values and results in a voxel-size dependence of R2*. Slice thickness is the largest contributor to voxel size, and has the largest contribution to the voxel-size dependence of R2* due to through-slice dephasing although the effects may be lower in the coronal orientation. While smaller and isotropic voxels are preferred, they are limited by the need for breath-holding. Adopting a fixed voxel size for all studies may be a preferred way to standardize the protocol and allow comparisons across sites. This is challenging to implement in routine practice because voxel size can be set directly on the scanners of one vendor (e.g. Philips), while in other vendors’ scanners (e.g. Siemens, GE) the voxel size is derived from field of view (FOV) and matrix size. Even if matrix size is kept constant, MRI technologists are inclined to change FOV based on the body habitus. Thus, standard operating procedures (SOPs) are necessary and the MRI technologist needs to be instructed not to change the FOV when prescribing renal BOLD MRI.
-
The presence of fat may have an effect on measured R2* values, and use of fat saturation or water-excitation pulses is preferred.
-
Most studies to-date have focused on the renal medulla which has a higher R2* (~30 s−1 at 3T) and hence a shorter T2* (~33 ms). In comparison, the renal cortex has R2* ~20 s−1 at 3T and T2*~50 ms. After fixing the longest TE = 50 ms, investigators need to decide the number of echo times. For a robust acquisition, any number of echoes between 8 and 16 would be acceptable. The actual number of echoes that is realizable will depend on readout bandwidth, echo spacing and image resolution. The bandwidth should be kept to around 300 Hz/pixel. With this echo time, use of either a mono-exponential or log-linear fit would be acceptable since the noise floor will not be reached.
-
R2* was the preferred metric and there was consensus for evaluating both cortical and medullary regions. ROI analysis was preferred, only because of widespread availability. While newer custom methods have advantages such as better objectivity, these also have several limitations. Until automated segmentation becomes routinely available, manual segmentation may be preferred.
-
There was an agreement that BOLD MRI provides a qualitative measure of relative oxygen availability and may be most suited for evaluating acute changes within the same kidney, e.g. following administration of furosemide. However, caution was advised when comparing different cohorts because of potential confounding effects such as blood volume fraction, oxy-Hb dissociation curve, and haematocrit [34]. Even though there was consensus in terms of reproducibility at a single site, similar data for multi-site studies remains lacking, and is thus desirable for future studies.
-
Similar recommendations are made for BOLD in native kidneys and allografts. However, the allografts’ positioning in the iliac fossa, closer to the skin during an end-expiration breath-hold, sometimes requires different mitigation strategies for the tissue-air interface susceptibility artefacts (e.g. placing a bag of saline on the skin over the expected region of the allograft before placing the imaging coil on the subject). Planning of slices in true coronal (to the allograft) orientation is somewhat more difficult than with native kidneys and in some studies an oblique sagittal slice with respect to the allograft’ longitudinal axis is preferred.
Discussion
Patient preparation
-
25% of respondents disagreed that diet should be controlled,
-
4% cited lack of publications showing an effect of diet on BOLD signal, and
-
4% thought that monitoring and reporting subjects’ diet is sufficient.
-
12.5% of respondents recommended fasting (water only intake) for 4–6 h before the scan, 8% went further and recommended abstaining from high protein foods 24 h before the scan in addition to fasting, to minimize susceptibility artifacts from bowel gas.
-
Another 8% of respondents thought restricting drinking before the scan can reduce artifacts from the bowel, and can have an effect on BOLD quality.
-
Another 8% of respondents recommended controlling sodium intake, possibly through standardized meals for study subjects in prospective studies.