November 13, 2000
Rockville, MD

Meeting Report
DRAFT COPY of March 19, 2001
(prepared by J. Evelhoch and S. Riederer)

• In the last several years contrast-enhanced MR angiography (CE-MRA) has become a commonly used, commercially available method. In parallel with this, dynamic contrast-enhanced MR imaging (DCE-MRI) has emerged as a promising method for diagnosis and prognosis of cancer. However, widespread use of DCE-MRI is limited by the need for further technical improvements.
• DCE-MRI and CE-MRA share the need for flexible acquisition methods to permit trade-off between temporal and spatial resolution. Hence, some of the considerable effort invested on CE-MRA acquisition methods should be translated to DCE-MRI.
• The focus of CE-MRA is on visualization of small vessels with demands for temporal resolution dictated by the desire to separate arteries and veins, while focus of DCE-MRI is on contrast kinetics with demands for spatial resolution dependent on the specific application. Consequently, translation will require consideration of these differing needs.
• Quantitative DCE-MRI, which is needed to permit its widespread use, requires special purpose data acquisition techniques and analysis software not normally included by the instrument manufacturers. Although precedents for development of advanced analysis package include functional neuro MRI (fMRI) and prostate MR spectroscopic imaging (MRSI), additional capabilities will be necessary to permit accurate generation of quantitative data.

The workshop brought together a limited number of researchers who are experienced in MRI pulse sequence techniques, image analysis, or applications in either CE-MRA or DCE-MRI with representatives from the MRI industry. The purpose was to identify desirable performance features and a possible framework of future pulse sequence and image analysis strategies which might be used for dynamic contrast-enhanced imaging of cancer to ensure that advances in this area can be widely implemented.

The program for the workshop is given in Appendix A. As shown, the program consisted of two parts, the first consisting of individual presentations on the meeting topic with several preceding presentations by NIH officials, and the second a discussion session. The list of all participants, including speakers and observers, is listed in Appendix B.

The workshop brought together MRI researchers who collectively represented a variety of backgrounds. This included individuals with a clinically based understanding of the requirements for DCE-MRI, others experienced in the physics-based development of MR acquisition techniques which have been used for CE-MRA, others with firsthand experience in DCE-MRI techniques as applied clinically to the imaging of cancer, and representatives from industry. Speakers were asked to present: (i) a review of their own work in CE-MRA or DCE-MRI with emphasis on the technical aspects and how they impact clinical utility; and (ii) a statement of their view of the future technical needs of DCE-MRI and how these might be met by innovations in acquisition and analysis methods. Detailed notes of the individual presentations and discussion as made on site by NIH staff are presented in Appendices C and D.

1. For DCE-MRI to be useful in a routine clinical setting, including clinical trials of new cancer therapies, acquisition and processing must be straightforward providing a 3D representation of the relevant quantitative information.
2. Properties of successful DCE-MRI methods include: cover entire tumor with best possible spatial and temporal resolution; flexible specification of FOV; results independent of system and contrast dose; free from spurious correlations; free from motion artifacts, free from errors due to tissue properties; include estimate of error; and highly reproducible.
3. Many CE-MRA techniques are relevant to DCE-MRI, but in general they are not directly transferable "off the shelf" from their current embodiments on commercial MRI machines.
4. Many meritorious technical directions are possible for attempting to get high 3D spatial and good temporal resolution. However, there is a need for more MRI sequence development both in general and for application specifically in DCE-MRI.
5. Most commercial MRI manufacturers already provide to their respective academic collaborators some level of capability to develop acquisition techniques of their own. In general this requires a technical knowledge of the MRI machine which is much deeper than that required for routine clinical operation.
6. Translation of new methods developed at one institution to other institutions for use in clinical trials of new cancer therapies requires personnel dedicated to supporting implementation at sites with considerably less technical knowledge of the MRI machine.
7. Requirements on temporal resolution may vary over the course of a contrast-enhanced examination, with resolution the order of seconds targeted for the initial contrast enhancement phase and becoming progressively longer to tens of seconds during later phases which may occur ten minutes later. This may vary according to the organ under study and is also subject to the preferences of the individual directing the examination.
8. A variety of the technical directions for research involve the acquisition of MR data in non-standard formats, using non-standard k-space trajectories and view orders, and with various ways to combine and filter data between different time frames as part of the image reconstruction. To study these effectively will require familiarity with details of the specific commercial MRI system used.
9. The application of DCE-MRI in an individual patient will likely require a greater degree of analysis than CE-MRA, as kinetic parameters must be estimated quantitatively. There is need for more research and development in this area of study.
10. Because of the manner in which k-space is sampled one trajectory (line, ray, or spiral) at a time, and because of the presence of time-varying signal levels in DCE-MRI, the very meanings of temporal and spatial resolution may need definition, and standard tests may need to be identified.
11. There is a need for parallel developments in humans as well as in animal models.
12. In order to have adequate patient statistics, the clinical study of a given new technique may well require the collaboration of investigators from multiple institutions. Recruitment of such collaborators may likely involve sites which focus primarily on clinical MRI vs. those with extensive technical research capability. Further, not all collaborators will necessarily be located near each other. In the short term, given the problems identified in #6, it may be best to specify minimum requirements, rather than specific sequence to facilitate cooperation from more clinically-oriented site.
13. The acquisition of high spatial resolution information over multiple timepoints will require the handling of large amounts of data. For example, a single contrast-enhanced 3D run with the order of 10-20 timepoints requires the order of several 100 MB of memory.
14. There are a variety of organs which appear appropriate for study by DCE-MRI. Although there is a common desire for high spatial and temporal resolution, there are organ-specific requirements as well. Nominal, targeted specifications on imaging parameters and contrast delivery for four specific organs are:

• Brain parameters: 2-3mm (cubed) isotropic voxels, whole brain FOV, T1 map in 1-2 minutes, 2 sec temporal resolution in first minute, slower out to 10 min. Measure of arterial function not affected by inflow. Fast bolus @3-5 ml/sec administration with saline chaser.
• Breast parameters: single or both whole breast FOVs; 10 sec temporal resolution at 2 mm cubed; with 0.5x 0.5x 3mm resolution in first 1-2 minutes; fat suppressed for CNR; run washout to 8 minutes; prior T1 map; 2-3 ml/sec contrast administration; measure arterial input function.
• Liver parameters: visualize >2cm lesions; FOV = 25 x 35 x 18 cm; continuous with acquisition-based motion correction; 5-10 sec temporal at spatial 2x2x7 mm spatial resolution; T1 map up front; 1-3 ml/sec bolus delivery; want arterial phase input function if feasible.
• Lung parameters: visualize >2cm lesions; FOV = 25 x 35 x 18 cm; continuous with acquisition-based motion correction; 5-10 sec temporal at spatial 2x2x5 mm spatial resolution; T1 map up front; 1-3 ml/sec bolus delivery; want arterial phase input function if feasible.

1. It is recommended that a research priority be the development of pulse sequences for magnetic resonance imaging of the contrast enhancement of organs for detection and characterization of cancer. The general goal is to provide high spatial resolution images of a volume containing the organ of interest and with temporal resolution adequate to characterize contrast pharmacokinetics. Among the possible areas of investigation are:
   • novel k-space trajectoreis (e.g. spiral, projection reconstruction, echo-planar, etc.)
   • novel ways for temporal sampling of k-space trajectories (e.g. phase encoding view order, view sharing, temporal filtering, interleaved spiral or projection reconstruction samples)
   • ways for characterization of the temporal response and spatial resolution of pulse sequences which can possibly be used for DCE-MRI (e.g. frequency-spatial resolution analysis, identification of standardized tests of spatial and temporal resolution appropriate for DCE-MRI)
   • means for dynamic adjustment of the tradeoff between spatial and temporal resolution during the course of an DCE-MRI run (e.g. dynamic adjustment of projection reconstruction sampling, dynamic adjustment of k-space sampling and extent of coverage)
   • new ways for providing improved spatial resolution per unit time as might be applied to DCE-MRI (e.g. multiple coil techniques such as sensitivity encoding "SENSE" and simultaneous acquisition of spatial harmonics "SMASH," effective utilization of high capability gradient technology)
2. It is recommended that a research priority be the quantitative evaluation of DCE-MRI time series of images, with development of techniques for determination of pharmacokinetic parameters characterizing tumor vascularity, with allowance for the potential application of these techniques using large data sets consisting of high resolution 3D volumes sampled at multiple timepoints.
3. It is recommended that a research priority be the application of current, emerging, and new DCE-MRI techniques in the imaging of cancer in humans for purposes of determination of the sensitivity, specificity, and utility of these techniques.
4. It is recommended that the NCI explore how it would encourage industry/academic partnerships to support and/or promote the effective distribution to less-experienced MRI sites of new methods developed by academic investigators for acquisition and quantitation of DCE-MRI time series images.
5. It is recommended that academic investigators with specific MR pulse sequences that meet, in part, the requirements for quantitative DCE-MRI acquisition in brain, breast, liver or lung (as described above) send their software or description to MRI industry to evaluate the ability of their system to implement the sequences.
6. It is recommended that a means should be explored for the major industry players to develop a process for evaluation of both acquisition and analysis software and make appropriate selections. (This may be problematic since they will each make their own decisions in their own way, with some influence from the direction others are taking).
7. It is recommended that a research priority be the development and study of animal models of cancer in which DCE-MRI may be evaluated

• Appendix A: Meeting Program
• Appendix B: List of Participants
• Appendix C: Notes by NIH Staff of Individual Presentations
• Appendix D: Notes by NIH Staff of Discussion Period