Is CT Perfusion Ready for Prime Time?

Depends on whom you ask. Technically, CT perfusion (CTP) of the brain is relatively easy to perform on the newest generation of ultrafast CT scanners. Generating meaningful images and data and then interpreting these data in a clinically relevant manner is a more difficult task. Two groups of neuroradiologists seem to have emerged: 1) the CTP cheerleaders, who believe CTP could (or even should) be a part of routine brain imaging, especially in the acute stroke setting, and 2) the CTP cynics, who are still waiting for more proof to be convinced that CTP will change the workup of patients with cerebrovascular disorders.

As occurs with most new neuroradiologic techniques, the avalanche of articles championing CTP has officially begun. Most of the early articles discussing and generally advocating this technique have appeared in vendor-supported journals (vendors like new techniques for obvious reasons) or as a part of “advanced imaging of stroke” review articles. (I, too, am not immune from writing such articles [1]). In attendance at the 2003 American Society of Neuroradiology meeting in Washington, DC, were neuroradiology luminaries preaching the gospel of CTP as a holy grail in the brain attack workup of stroke patients. In addition to review articles, scientific studies of CTP have begun to appear in peer-reviewed journals, and several groups have begun to employ CTP on a routine clinical basis. These groups have reported their preliminary sensitivity and specificity of CTP compared with other techniques, such as MR diffusion and perfusion (2), and have attempted to establish the clinical utility of this technique (3, 4). In addition, methodological advances are creeping into the literature (5). These studies usually have high physician and technologist person-hours invested into creating quality CTP images with highly trained people interacting with the data. It is with this backdrop that Fiorella et al’s article “Assessment of the Reproducibility of Postprocessing Dynamic CT Perfusion Data” appears in the current issue of the AJNR.

Fiorella et al take a slightly different, yet very practical, approach to evaluating this new technique. Rather than looking at this technique with the best possible product achievable and then evaluating the utility of the technique, they took a stab at evaluating the clinical practicality of CTP. Specifically, the authors aimed to assess the reproducibility of the postprocessing step in the analysis of CTP data in a group of 20 patients. Three technologists measured four variables: arterial input function (AIF), venous function (VF), preenhancement interval (PREI), and postenhancement image (POEI). The other potential variables, such as region of interest location for the cerebral blood flow (CBF), cerebral blood volume (CBV), and mean transit time (MTT) measurements, were removed by the fact that identically sized predetermined locations were already chosen by a neuroradiologist. Despite a “technologist outlier,” the authors showed high intraclass and interclass correlations for the calculated CBV, CBF, and MTT values. Also, the AIFs, VFs, and PREIs chosen by the technologists were similar. The POEIs, however, were often markedly different, leading to differences in the appearance of the CTP maps. Thus, the within-subject standard deviations actually showed clinically significant levels of variability within the maps. The authors, therefore, conclude that, although there is high correlation between region-of-interest measurements in the CTP maps, the degree of agreement may not be sufficient to allow the use of quantitative values in the clinical practice, but that postprocessing parameter optimization will naturally reduce the variability observed in this study.

Of course, such conclusions have important implications to our clinical practice. This study attempts to mimic everyday practice in which a technologist will undoubtedly be the interface between the data acquisition and image analysis. In most clinical centers, it is just not practical to have only the neuroradiologists perform all of the data manipulations. Fiorella et al know this and appropriately remove the neuroradiologist from the postprocessing step. They point out that the postprocessing techniques must be reproducible across technologists in order for the technique to have clinical practicality.

So the question remains: is CTP ready for prime time? Without question, CTP provides the clinician with more knowledge about the underlying perfusion and cerebrovascular status of the patient. The idea of using perfusion as a tool to evaluate the brain is not a new concept. It really got off the ground with MR imaging, first by using bolus gadolinium injection and later arterial spin labeling techniques. In fact, diffusion-perfusion mismatches were touted several years ago as possibly being the diagnostic tool that identifies the holy grail of stroke imaging—the ischemic penumbra. Although it was possible to perform MR perfusion relatively easily, it was (and is) still not clear which perfusion images should be evaluated: the blood flow images, the blood volume images, or the MTT images. Also, time to peak and other summary parameter images could be created. Another question also arose about whether quantified cerebral blood maps were necessary or whether qualitative maps suffice. Such questions were never fully answered when CTP techniques emerged and allowed a more practical technique in the emergency room setting.

CT affords the opportunity of one-stop shopping in the emergency room setting, and because “time is brain” in the acute stroke setting. Practically speaking, CT definitely has an advantage over MR imaging. A routine CT, CT angiography, CTP protocol may ultimately prove to be the protocol in most centers, and such a protocol would allow a rapid assessment of the anatomy, vascular patency, and regional hemodynamic status on an individual basis. Thus, one can easily envision a CT protocol that might guide clinicians in the early stages of an infarct rather than simply basing decisions on population studies. As already mentioned, some centers are already using such a protocol to stratify patients into different treatments.

As Fiorella et al show in their study, however, there are some inherent problems with the CTP technique, and each facility must attempt to attain uniformity in performing these studies. Their article represents a necessary step along the pathway of critically evaluating CTP as a potential diagnostic tool.

• Copyright © American Society of Neuroradiology