A spate of recent articles, including one by Saur et al in the May 2003 issue of AJNR (1), has confirmed the “obvious:” namely, that diffusion-weighted MR imaging is more sensitive and has greater interrater agreement than unenhanced CT for the detection of early ischemic signs of stroke.
Confirming the “obvious” is not an unimportant or trivial task and can sometimes lead to unexpected results. A possibly apocryphal legend has it that up until the time of the Renaissance when the Italian anatomist Andreas Vesalius actually looked in a horse’s mouth to verify for himself what he would find there, textbooks incorrectly described the number of teeth horses have on the basis of centuries-old authority of the ancient Roman physician Galen. Analogously, in one of the first clinical reports of the diagnostic accuracy of diffusion-weighted MR imaging in acute stroke detection, Gonzalez et al (2) surprised us not so much with the finding that diffusion-weighted imaging has greater sensitivity than that of unenhanced CT (100% versus 45% in the small patient cohort studied), but with the revelation that in blinded review the sensitivity of unenhanced CT far exceeds that of conventional T2-weighted and proton density-weighted MR imaging (45% [unenhanced CT] versus 18% [T2- and proton density-weighted MR]). Indeed, until the advent of thrombolytic agents, which proved to be of benefit for acute stroke victims, and the consequent widespread need for imaging triage, the often-subtle unenhanced CT signs of early ischemia had typically been both overlooked by clinicians and underreported in the literature. Such early ischemic signs include: 1) parenchymal hypoattenuation with loss of gray matter-white matter differentiation owing to cytotoxic edema and possibly decreased blood volume (eg, “insular ribbon” sign); 2) sulcal effacement, also owing to edema; and 3) hyperattenuated vessels owing to intraluminal thrombus (eg, “hyperdense middle cerebral artery [MCA]” sign).
Saur et al considered all of these factors in their CT assessment of early ischemic changes, with a resultant sensitivity of 73% (versus 93% for diffusion-weighted imaging) based on the consensus ratings of three neurologists, and 87% (versus 98% for diffusion-weighted imaging) based on the consensus ratings of three neuroradiologists (P = .04 for neurologist versus radiologist CT interpretation and P = .30 [NS] for neurologist versus radiologist diffusion-weighted imaging interpretation). These results are novel and noteworthy because, as the authors point out, earlier studies comparing CT and diffusion-weighted imaging findings were confounded by the relatively long time interval between the admission CT and initial diffusion-weighted examinations. In this investigation, the mean delay between imaging sessions was a brief 25 minutes. The authors’ conclusion that diffusion-weighted imaging depicts early ischemia with higher sensitivity than that of CT has received strong recent confirmation. Fiebach et al (3) clearly showed this in a study in which CT and diffusion-weighted images were randomly obtained. That the radiologists performed significantly better than the neurologists for CT, but not for diffusion-weighted imaging, supports the contention that interpretation of subtle stroke CT findings is a learnable skill that improves with experience, but that interpretation of highly conspicuous diffusion-weighted imaging findings requires little specialized training. To be sure, arguably the greatest benefit of using an objective CT grading scheme, such as the Alberta Stroke Program Early CT Score ([ASPECTS]), for which interrater agreement is superior to that of the “1/3 MCA” rule, is that it compels the inexperienced reader to carefully examine all portions of the CT image (4).
The remarkably high sensitivity for acute stroke detection achieved by the neuroradiologists in this study (approaching 90%) is noteworthy and is likely related to the specific population studied, which consisted predominantly of large-vessel embolic stroke patients. Also, each reader was aware of the global suspicion for stroke during image analysis, which may explain why not blinding to the clinical history did not alter the results. Care was taken to optimize both imaging technique and image interpretation; center level and window width settings of the hardcopy CT images were appropriate for the detection of subtle decreases in Hounsfield attenuation. If anything, the tube current (mA) used during scan acquisition was larger than what is minimally required for an adequate signal-to-noise ratio (voltage was not reported, but is assumed to be 120–140 kV). The breakdown by time-to-imaging of the CT and diffusion-weighted sensitivities for detection of early ischemic signs, shown in Table 1 of the article by Saur et al, not only underscores the importance of time as a critical determinant of infarct conspicuity, but serves as a reminder that different pathophysiologic phenomena underlie the acute CT and diffusion-weighted imaging findings. Indeed, one wonders from this data if the sensitivity of CT and diffusion-weighted imaging are really all that different beyond a 3–4 hour time window.
Finally and most importantly, Saur et al’s conclusion that their results “support the application of ‘stroke MR imaging’ for the management of acute stroke patients” fails to take into account the evolving use of contrast-enhanced CT techniques for neurovascular evaluation. Because, as compared with MR imaging, CT is rapid, inexpensive, and more readily available in a variety of urgent care settings, there is strong current interest in developing a combined unenhanced CT, CT angiography, and CT perfusion protocol for thrombolysis triage. Preliminary studies from multiple groups, including our own, suggest that the sensitivity of postcontrast CT angiography source images for acute stroke detection approaches that of diffusion-weighted imaging for all but the smallest distal emboli and lacunar infarcts (5). Moreover, there is increasing evidence from the MR, CT, and nuclear medicine literature that it is the degree, and not simply the volume, of ischemic change on blood volume and blood flow maps that may be a critical determinant of clinical and imaging outcome, as well as hemorrhagic risk, in response to thrombolysis. Thus, in the ongoing battle between CT and MR imaging as the first-line technique for acute stroke imaging, the “obvious” choice may not necessarily prove to be the correct one. Stay tuned.
References
- Copyright © American Society of Neuroradiology