J. Evelhoch^a, T. Brown^b, T. Chenevert^c, L. Clarke^d, B. Daniel^e, H. Degani^f, N. Hylton^g, M. Knopp^h, J. Koutcherⁱ, T.-Y. Lee^j, N. Mayr^k, D. Sullivan^d, J. Taylor^l, P. Tofts^m, R. Weisskoff^n
aWayne State University, Detroit, MI; ^bFox Chase Cancer Center, Philadelphia, PA; ^cUniversity of Michigan, Ann Arbor, MI; ^dNational Cancer Institute, Bethesda, MD; ^eStanford University, Palo Alto, CA; ^fWeizmann Institute, Rehovot, Israel; ^gUniversity of California, San Francisco, CA; ^hGerman Cancer Research Center, Heidelberg, Germany; ⁱMemorial Sloan Kettering, New York, NY; ^jRobarts Research Institute, London, Ontario; ^kUniversity of Iowa, Iowa City, IA; ^lSt. Jude Children's Research Hospital, Memphis, TN; ^mUniversity College London, London, England; ⁿEPIX Medical, Cambridge, MA

Purpose

Establish minimum requirements for standardized data acquisition for oncologic applications of dynamic contrast enhanced MRI (DCE-MRI) to allow integration of data from different institutions and comparison of various approaches for data analysis.

Introduction

DCE-MRI has recently emerged as a promising method for both diagnosis and prognosis of cancer. Remarkably, these positive results have been obtained despite considerable variation in both the methods of data acquisition (e.g., pulse sequences, acquisition parameters, temporal resolution, spatial resolution and coverage) and analysis (e.g., visual inspection , parametric analysis , pharmacokinetic or physiologic modeling). This suggests there are substantial physiologic/pharmacokinetic differences (i.e., between benign and malignant, or between non-responsive and responsive tumors) underlying these observations that are evident independent of the methods used for acquisition and analysis of the DCE-MRI data. Moreover, these encouraging results support a potential role for DCE-MRI in the development of a new class of anti-cancer agents based on action against tumor angiogenesis . Clearly, there is a promising future for use of DCE-MRI as both a clinical research tool and in routine clinical practice. However, several issues must be addressed to fulfill this promise.

The fundamental issue impeding realization of this promise is that integration of results from multiple institutions and/or evaluation of the relative merits of the various methods for data analysis are difficult, if not impossible, in most cases. This is due to the variety of methods used for data acquisition and analysis and the lack of a general consensus concerning how best to acquire and/or analyze DCE-MRI data. Consequently, the relative clinical relevance of the information provided by these different approaches is not known. Several factors could impact the information derived from DCE-MRI data (e.g., intra- or inter-patient variation in either the initial T1 or the blood contrast agent concentration as a function of time - the 'arterial input function'). Although the importance of each of these factors for clinical research tools and routine clinical practice may differ depending on the specific application, it would be useful to establish baseline requirements for general research and/or clinical studies. Also, it is desirable to maintain high spatial resolution as to not compromise morphology-based interpretations. This would facilitate both integration of data from different institutions and comparison of various approaches for analysis of the kinetic data.

Recommendations

The following recommendations represent a consensus reached at an NCI-sponsored workshop held in October 1999 to address these issues for low molecular weight Gd-based extracellular contrast agents. These recommendations are an independent statement of the workshop participants and do not represent a policy statement of the NIH or Federal Government.

• Pre-injection

• If possible, measure T1 (using same resolution and field of view for dynamic data)
• Acquire maximum spatial resolution image (determined by application)

• Contrast agent injection

• If possible, use power injector to minimize variation
• 15-30 sec for total injection, saline flush

• Dynamic study

• If possible, sample arterial input function
• For first 90-150 sec after bolus injection, use 10-30 sec temporal resolution (fastest sampling possible consistent with spatial resolution requirements)
• Acquire centric phase-encoded higher spatial resolution images out to 10 min with 1-4 min temporal resolution

Discussion

The rationale is to record potentially rapid signal changes as they occur (albeit at reduced spatial resolution), then transition to high spatial resolution imaging as the intensity changes become less abrupt. Use of the recommended 'adaptive imaging' approach would facilitate both comparisons among different groups and evaluation of the merits of the various approaches available for data analysis. We recognize that all interested investigators may not be able to implement the recommended 'adaptive imaging' approach immediately. This is primarily for two reasons: (1) 'adaptive imaging' and input function sampling requires very flexible control of clinical MR scanners to permit dynamic switching between high spatial and temporal resolution; and (2) methods optimizing the number of pixels imaged per unit time may not be available on all systems. Increased cooperation among investigators, manufacturers and the relevant government agencies should facilitate more widespread use of the recommendations.

References

1. CA Hulka et al., Radiology 205, 837-42 (1997)
2. CK Kuhl et al., Radiology 211, 101-10 (1999)
3. H Hawighorst et al., Clin Cancer Res 4, 2305-12 (1998)
4. NA Mayr et al., AJR Am J Roentgenol 170, 177-82 (1998)
5. Reddick et al., JMRI 10, 277-285 (1999)
6. T Boehm et al., Nature 390, 404-7 (1997)
7. Beauregard DA et al., Br J Cancer 77, 1761-7 (1998)