Bleed or Stroke? Diffusion Measurements in Intracranial Hematomas

In the brief time that it has been readily available in the clinical arena, diffusion-weighted imaging (DWI) has become an integral part of many protocols, notably those directed at imaging the patient suspected of harboring an acute ischemic infarct (1). As experience has been gained, however, it has become clear that not all lesions that reduce diffusion are infarcts. Indeed, reduced diffusion has been reported in acute encephalitis, in acute demyelinating diseases such as multiple sclerosis, in abscesses, and in lymphomas. Calling everything that reduces diffusion an acute infarct is a pitfall to be avoided.

At the same time that technical advances in MR imaging have made diffusion imaging fast and widely available, advances in acute stroke therapy have mandated that imaging studies be performed and interpreted rapidly. To this end there has been a move in some institutions toward immediate MR imaging in the setting of suspected acute ischemic infarction, rather than the traditional performance of the noncontrast CT to assess for intracranial hemorrhage or other processes that may mimic acute ischemic infarction clinically. Therefore, if the patient suspected of having an acute infarction undergoes only MR imaging, with an abbreviated protocol that may include only diffusion and perfusion imaging, possibly with fluid-attenuated inversion recovery (FLAIR) and MR angiography, a thorough understanding of the appearance of blood products is necessary to interpret appropriately the diffusion images. In this issue of the AJNR, Atlas et al (page 1190) report the appearance of 17 intracranial hematomas on both conventional and diffusion-sensitive MR images, and discuss how this information furthers our understanding of the evolution of signal characteristics of hematomas over time.

Sixteen consecutive patients with 17 intracranial hematomas (proven by CT, surgery, or both, and not related to tumor, infarction, or trauma) were studied with T1- and fast spin-echo T2-weighted imaging, as well as DWI from which apparent diffusion coefficient (ADC) maps were calculated. All phases of hematoma evolution were represented based on conventional MR imaging criteria. The authors found that the hematomas could be segregated clearly into two groups based on their average ADC values. Those hematomas containing intact red blood cells (eg, hyperacute, acute, and early subacute hematomas) had significantly reduced diffusion compared to those containing lysed cells (subacute to chronic hematomas). Also of note, the ADC values of all early hematomas were reduced significantly compared to normal white matter. Potential causes offered to explain this phenomenon included: a decrease in volume of the extracellular space with clot retraction, a change in the osmotic environment of extravascular blood such that the shape of the red blood cell is altered, formation of the fibrin network associated with clot, conformational changes of the hemoglobin molecule, or a contraction of intact red blood cells with a decrease in intracellular space. There is evidence that the relative size of intracellular and extracellular compartments may influence the appearance of ischemic infarcts on DWI significantly (2), and this is probably relevant to the diffusion properties of hematomas as well. Relative contributions of this and other mechanisms remain to be investigated in the future.

Intracranial hematomas change over time in many ways, two of which are particularly important: first, the oxygenation state of hemoglobin changes, and second, red blood cells lyse. During the hyperacute, acute, and early subacute phases of hematoma evolution, hemoglobin is oxygenated initially, then undergoes deoxygenation, and finally becomes oxidized and forms methemoglobin; all of this occurs within an intact red blood cell. In the late subacute to chronic phases of hematoma evolution, the red blood cell membrane lyses and methemoglobin becomes extracellular. Hemoglobin in its various states has variable magnetic susceptibility effects, which contribute significantly to the appearance of a hematoma of a given stage on conventional MR images. Diffusion-weighted MR sequences are primarily sensitive to changes in water motion, a parameter that is influenced by a number of factors including: the relative size of the intracellular and extracellular spaces, the presence or absence of intact cell membranes, and the degree of anisotropy of a given tissue. One might therefore expect that the presence or absence of intact red blood cell membranes would influence significantly the appearance of an intracranial hematoma on diffusion-weighted images. Of course, one must work within the caveat that the diffusion-weighted image is many things, ie, it is not only sensitive to microscopic motion of water, but it is also sensitive to T2 and magnetic susceptibility effects due to its long echo time and echo-planar acquisition. Furthermore, the postprocessing of diffusion-weighted images to obtain ADC maps presupposes the presence of some signal on the images. We have observed cases where acute hematomas, presumably composed of deoxyhemoglobin, have appeared purely as signal voids on T2- and diffusion-weighted images, and so the interpretation of postprocessed diffusion attenuation images and ADC maps does not yield meaningful information with regard to diffusion properties. This problem is shown in part by Figure 2 (page 1193) in the article by Atlas et al; the black rim around the hematoma on the diffusion-weighted image presumably is due to susceptibility effects and does not yield useful information on the ADC map, whereas the bright center corresponds to a region of reduced diffusion and is portrayed accurately on the ADC map.

The data presented in this paper suggest that early hematomas (containing intracellular oxyhemoglobin, deoxyhemoglobin, or methemoglobin within intact red blood cells) could appear identical to the signal intensity of acute infarction on diffusion-weighted images and ADC maps despite their clear differentiation on conventional MR images. Therefore, diffusion-weighted images and ADC maps obtained in the context of acute neurologic deficit always should be interpreted in the context of conventional MR images (at the very least the echo-planar T2-weighted image that is obtained as part of the diffusion-weighted sequence with b = 0). We would agree with this concern and would add the further caveat that one must review the diffusion-weighted image carefully and not simply draw conclusions from the ADC map alone. In the same way that diffusion-weighted images can be ambiguous on the basis of “T2 shine-through” (3), ADC maps can be ambiguous if they are derived from diffusion-weighted images that lack signal completely due to, for example, magnetic susceptibility effects. Our acute stroke protocol at present includes sagittal T1-weighted images, axial FLAIR images, diffusion and perfusion imaging, intracranial MR angiography, and postcontrast T1-weighted imaging, with the latter two eliminated in the uncooperative patient. In addition to viewing the source of diffusion-weighted images, we perform postprocessing to obtain diffusion attenuation images and ADC maps. We agree with Atlas et al's note of caution with regard to the potential for misinterpretation of diffusion imaging in the context of acute neurologic deficit, and echo the notion that data from each sequence must be incorporated into a coherent clinical and imaging picture.