Dynamic susceptibility contrast (DSC) magnetic resonance imaging (MRI) of microvascular hemodynamic parameters has the potential for guiding the development of new drugs or devices, as well as improving diagnosis and disease staging for individual patients. For example, precise measurement of these parameters in individual patients and controls may be useful in developing antineogenic (i.e., antineoplastic) agents. Such measurements can also be useful for assessing tissue viability and the efficacy of reperfusion maneuvers used for acute ischemic stroke. A number of multisite clinical trials and multisite observational studies have already used DSC MRI.^{1,2,3,4,5,6,7} Given that most modern-day clinical trials are performed across many sites to optimize statistical power within a practical time period, realizing the full potential of DSC MRI hemodynamic parameter imaging requires a means of establishing whether different sites produce biased hemodynamic measures and, if so, developing a means of correcting for the biases. Furthermore, reproducible measurements across sites might enable the use of threshold values (either directly applied to perfusion maps or after calculating ratio maps by comparing to the contralateral hemisphere or reference area) in the diagnostic process for staging of disease severity (e.g., tumor grade). Beyond the scope of clinical trial research and everyday patient care, it is also important to consider that advancing knowledge in any medical field is enhanced by reliable and reproducible measurements that can be used with confidence at any site. Without between-site consistency, the medical literature can become confusing, as small studies done at individual sites may become contradictory.

DSC MRI is a complex procedure. This is underscored by Willats and Calamante’s⁸ description of no fewer than 39 issues to be aware of in the collection of and processing of DSC MRI data. Complex procedures are more likely to produce site-specific biases. Site specific bias in DSC MRI measurements can be attributed to the following causes:

A meta-analysis was performed to evaluate the degree of variability in DSC image acquisition techniques. Publications that described use of DSC MRI for clinical research purposes that were published in the English language in 2011 and 2012 and were available through the first author’s library were included. The keywords “dynamic susceptibility contrast magnetic resonance imaging” were used in a Medline search to identify candidate publications. Technical development studies were excluded to avoid confusion of emerging methods with established ones that are currently in common clinical use. Key parameters obtained from the methods sections of these publications appear in Tables 32.1 and 32.2. These data demonstrate the degree of variability that is to be expected in a hypothetical clinical trial if no prospective effort was made to establish a homogeneous across-site DSC MRI acquisition procedure.

Several key summary points emerge from the data shown in Tables 32.1 and 32.2. Perhaps the most interesting is that some authors manage to publish without specifying any details about the DSC MRI methodology they used. All but one of the studies listed in Tables 32.1 and 32.2 is focused on brain imaging. This may be an artifact of the keywords used in the Medline search or it may be an artifact of the time span used for the search. If not an artifact, it suggests that DSC MRI is not being used for research purposes related to body imaging or cardiac imaging, which might be explained by the presence of the blood–brain barrier, making DSC MRI very suitable for neurologic applications. Outside the neuroaxis, contrast agent leakage and the challenges of performing high-quality EPI scans has mostly favored DCE-type acquisitions, which are typically based on T₁-weighted three-dimensional sequences and pharmacokinetic modeling. Scanner hardware in use at clinical research sites is almost equally distributed between 1.5T and 3T for these brain studies. GE and Siemens MRI units are being used in about equal proportion, with the use of Philips scanners being about half of the use of GE or Siemens scanners. The majority of reports define the acquisition pulse sequences as GE EPI or with terminology that can be regarded as the equivalent (e.g., fast-field echo EPI). The outliers are several studies that used principles of echo shifting with a train of readouts (PRESTO) pulse sequences with Philips scanners. Volume image acquisition time resolution varied from 600 to 3000 milliseconds. The number of volumes in the time series that were acquired varied from 10 to 120. The average echo time (TE) used at 1.5T was 45 ± 18 milliseconds (mean ± standard deviation) and the average TE used at 3T was 35 ± 10 milliseconds. In many instances, flip angle was not reported. If the PRESTO studies are excluded because PRESTO uses a significantly different signal excitation method, the flip angle used varied between 30 and 90 degrees and was only weakly correlated with the time used to collect one stack of



two-dimensional slices (i.e., repetition time [TR]). In-plane spatial resolution tended to be about 2 mm but also showed some variability. Slice thickness (through-plane resolution) varied between 3 and 7.5 mm.