Diffusion Findings in Blood Clot: The Last Word?

Diffusion-weighted imaging (DWI) is most recognized for its diagnostic utility in stroke; however, recent attention has focused on other diseases that similarly exhibit restricted diffusion on DWI. DWI of blood clot is of particular interest because hemorrhage may complicate the appearance of stroke. Because the process of clotting involves transformation of a fluid to a semisolid, it is predictable that water diffusion would decrease in acute clot and hence be hyperintense on DWI. Thus, the AJNR articles by Atlas et al (1) and Maldjian et al (2) focusing on the diffusion characteristics of intracerebral hematomas are of interest in that they represented early, albeit incomplete, studies of the diffusion characteristics of blood clot as a function of time.

The earlier study, by Atlas et al, did not address the appearance of blood clot on diffusion-weighted images, restricting the focus to apparent diffusion coefficient (ADC) values. They reported significantly reduced ADCs compared with normal white matter in early hematomas (hyperacute, acute, and early subacute) but increased ADCs after cell lysis occurred—ie, in the late subacute phase (1). The finding of reduced ADC in clot was contested by Maldjian et al, who argued that automated ADC calculations may be underestimated when using vendor-supplied software because of thresholding effects at low signal intensity-to-noise (SNR) ratios. They argued that ADC was not restricted in phases of clot in which red cells are intact, concluding that those clots have essentially the same ADC as white matter when proper technique is employed. Although the mean ADC of the 12 hematomas in the study by Maldjian et al was not significantly different from white matter, two of the four hyperacute hematomas did have markedly decreased ADC, and late subacute clots were not studied at all. Thus, these two studies left unanswered questions regarding the DWI appearance and ADCs of intracranial blood clots. Hence, we sought to readdress the issue in a larger group.

Before we completed collecting our cases, however, we found a more recent and thoroughly performed study by Kang et al (3) whose results are so consistent with our own that we stopped the study early. The Kang et al study found that clots were bright on DWI in hyperacute and late subacute clots and that ADCs were reduced compared with normal brain tissue during all phases (hyperacute, acute, and early and late subacute). These data suggest that diffusion is restricted in clots before and after cell lysis, resulting in bright signal intensity on DWI unless T2 effects of intracellular unpaired electrons reduces the signal intensity (SI) (T2 dark-through effect [2]).

Our data mirror those of Kang et al. Twenty clots (hyperacute [n = 3], acute [n = 7], early subacute [n = 5], late subacute [n = 5]) were studied on T1-weighted images, T2-weighted images, DWI, and ADC maps, and the results were expressed as SI ratios (Fig). Hyperacute clot was markedly hyperintense on DWI in three of three cases; acute clots were markedly hypointense in seven of seven cases. In four of five early subacute clots, DWI SI was hypointense. In five of five late subacute clots, DWI was hyperintense. To address the concern of Maldjian et al regarding thresholding effects by using vendor-supplied software, we calculated ADCs by using 0% and 20% thresholds, and it made no significant difference in the appearance of ADC maps. In addition, for quantitative study (Fig), we recorded region of interest SIs and manually calculated ADCs by using the Stejskal-Tanner formula (4). We observed the marked hyperintensity of hyperacute clot to be associated with restricted diffusion (low ADC). The conspicuous hypointensity of acute and early subacute clots was also associated with low ADC, concurrent with marked T2 effects dominating SI. In late subacute clots, SI returned to a hyperintense appearance on DWI as T2 effects dissipated and restricted diffusion persisted.

In conclusion, our data concur with those of Kang et al in indicating that diffusion is reduced in hyperacute, acute, and subacute clots. Reduced ADC accounts for the marked hyperintensity on DWI scans in hyperacute and late subacute phases. Despite restricted diffusion, SI on DWI is not increased in the intervening acute and early subacute phases because of T2-induced hypointensity of clot, which dominates signal intensity on DWI (ie, “T2 dark-through”).

References

Atlas SW, Dubois P, Singer MB, Dongfeng L. Diffusion measurements in intracranial hematomas: implications for MR imaging of acute stroke. AJNR Am J Neuroradiol 2000;21:1190–1194
Maldjian JA, Listerud J, Moonis G, Siddiqi F. Computing diffusion rates in T2-dark hematomas and areas of low T2 signal. AJNR Am J Neuroradiol 2001;22:112–118
Kang BK, Na DG, Ryoo JW, et al. Diffusion-weighted MR imaging of intracerebral hemorrhage. Korean J Radiol 2001;2:183–191
Stejskal ED, Tanner JE. Spin diffusion measurements: spin echoes in the presence of time dependent field gradient. J Chem Phys 1965;42:288–292

Reply

Diffusion Findings in Blood Clot: The Last Word? Not!

Shah and colleagues report that hematomas have slow diffusion in the hyperacute, acute, and early subacute stages, findings similar to those of Kang et al (1). The apparent diffusion coefficient (ADC) values of late, subacute hematomas are less clear in Shah et al’s study, but the authors imply that it also was slower than those of normal white matter. We suspect that these results are classic examples of the artifact reported by Maldjian et al (2).

As previously described, standard vendor-supplied software does not produce valid signal intensity measurements in regions of extremely low signal (2). The attempts, first by Atlas et al (3), then by Kang et al (1), and now by Shah and colleagues, to report measurements from an area of signal void remain confounded by measurement procedures. The pitfall of susceptibility induced signal losses complicating ADC measurements in acute hematomas was even raised in an editorial published concurrently with the Atlas article (4). The purpose of our study was to provide a framework for computing ADC values in the setting of low-T2 signal (2). Neither Kang et al nor Shah and colleagues letter take this problem into account. Specifically, in acute and subacute hemorrhage, the T2*-weighted signal intensity can be close to that of background noise. This low signal intensity can result in acute hematomas appearing dark or black on diffusion-weighted images and ADC maps (T2 blackout). Obtaining an accurate diffusion measurement is problematic in this setting, because an individual pixel value may be dominated by the thermal and electronic noise of the imaging system. In the presence of background variations, it is even possible for a pixel at background intensity to show a higher magnitude of signal intensity on the diffusion-weighted image than on the baseline image, producing a spuriously negative ADC value (which is nonsensical and a violation of the second law of thermodynamics). Inclusion of such pixels in a region of interest (ROI) will lead to artifactually low mean ADC values. In our article, we provide a framework by using Expected Values to compute ADC values in this setting. We demonstrate that using vendor-supplied software that automatically masks the background (intended to provide more visually pleasing results) is an inherently flawed method of computing ADC values in these cases. Even at low levels of background masking (2%), the resulting ADC maps show marked artificial decreases in values from masking out areas of low signal intensity (including these masked values in the ROI). Apparently, the Kang study (1) did not take this into account when generating ADC measurements. In fact, it is unclear how they were generated. Kang et al do not provide an ADC map for a visual assessment of their methodology.

In their study, Shah and colleagues state that there were no differences in ADC maps between 0 and 20% thresholds. It is difficult to know what this means. Systematically and selectively removing a large portion of data, those pixels with the lowest signal intensity would have to change the results. Absent noise, this exclusion would select against pixels with the fastest diffusion and produce artifactually low ADC values. If it did not, then something must be seriously wrong with the pulse sequence or the measurement. The authors report that calculating ADCs by using 0 and 20% thresholds made no difference in the appearance of ADC maps. It is not clear whether Shah and colleagues measured ADC values at both 0 and 20% thresholds and found no differences. If not, we do not know what to conclude from the statement that the maps appeared similar.

We have shown that at levels of background masking as low as 2%, significant numbers of dropped pixels are evident in ADC maps of acute hematomas (2). Did Shah and colleagues find no such differences? Acute hematomas demonstrate T2* signal intensity close to that of background noise. If T2 black hematomas had signal intensity >20% of the mean brain signal, they would not be T2-dark. With some vendor-supplied software, there may be masking present in the algorithm, even at the 0% nominal masking. It is also possible that there was no signal because of susceptibility effects, even at the 0% masking.

Shah and colleagues state that their diffusion-weighted images were markedly hypointense (ie, black) for all the acute hematomas and hypointense for four of the five early subacute hematomas. That these hematomas were very dark on the diffusion-weighted images indicates that an accurate diffusion measurement cannot be obtained without accounting for the masking effect.

Atlas et al have observed that ADC is reduced on average in the acute phase of a hematoma, as compared with white matter, and has suggested that this is due to restricted diffusion (3). Although we would agree that restricted diffusion plays a role as a lower limit (2), we argue that variability in the measured ADC is dominated by the amount of extracellular fluid present, a quantity that we believe to be highly variable. In addition, the diffusion measurements made using the methods of Atlas et al, Shah et al, and Kang et al on T2-dark hematomas are all suspect. Atlas and colleagues further observe that lysis of the red blood cells will increase the ADC as compared with white matter; we have no reservations regarding this observation.

Shah and colleagues, on the other hand, state that the ADC is reduced in late subacute clots. From this observation, Shah et al conclude that “restricted diffusion persisted.” “Slow diffusion” and “restriction” are not synonymous. Restriction is one of several potential causes of slow diffusion. The finding of slow diffusion by itself does not permit a conclusion that restriction is the mechanism. The finding of slow diffusion after cell lysis, in which there are no apparent barriers to produce restriction, would argue against restriction as a plausible mechanism.

Claims for hyperacute hematoma are also confusing. Atlas et al make no claim, whereas Kang et al and Shah et al report a mean ADC ratio of approximately 0.7. Given that the hyperacute hematoma most resembles fresh blood, its ADC would presumably also resemble that of plasma. This is additionally supported by a recent article by Wintermark et al (5) in which diffusion is increased in hyperacute hematomas, in contrast to the assertions of Shah and colleagues.

What is the last word on diffusion in blood clot? We don’t know; but establishing the presence of restriction in hematomas would require a set of experiments that are rarely performed with clinical instruments. In order to determine the true diffusion signal characteristics of hematomas by use of MR imaging, it will likely be necessary to implement pulse sequences similar to that of the recently described propeller fast spin-echo technique (6), which are less prone to the susceptibility induced dephasing and distortions inherent in echo planar-based methods. Additionally, the relationship between blood susceptibility and ADC may be a multiexponential (7, 8) rather than a single exponential decay model assumed in the standard Stejskal-Tanner relationship (9). In fact, it is unlikely that a clear answer will be determined by using in-vivo human data, as the extracellular fluid fraction in intracranial hematomas may often be the dominant contributor to diffusion rates and can vary from patient to patient. It is clear, however, that the diffusion-weighted signal intensity for blood products is complicated, and measurements derived from areas of signal void and computed with incompletely documented algorithms will be of limited value.

References

Kang BK, Na DG, Ryoo JW, et al. Diffusion-weighted MR imaging of intracerebral hemorrhage. Korean J Radiol 2001;2:183–191
Maldjian JA, Listerud J, Moonis G, et al. Computing diffusion rates in t2-dark hematomas and areas of low T2 signal. AJNR Am J Neuroradiol 2001;22:112–118
Atlas SW, DuBois P, Singer MB, et al. Diffusion measurements in intracranial hematomas: implications for MR imaging of acute stroke. AJNR Am J Neuroradiol 2000;21:1190–1194
Fischbein NJ, Roberts TP, Dillon WP. Bleed or stroke? Diffusion measurements in intracranial hematomas. AJNR Am J Neuroradiol 2000;21:1179–118
Wintermark M, Maeder P, Reichhart M, et al. MR pattern of hyperacute cerebral hemorrhage. J Magn Reson Imaging 2002;15:705–709
Pipe JG, Farthing VG, Forbes KP. Multishot diffusion-weighted FSE using propeller MRI. Magn Reson Med 2002;47:42–52
Zhong J, Kennan RP, Gore JC. Effects of susceptibility variations on nmr measurements of diffusion. J Magn Reson 1991;95:267–28
Does MD, Zhong J, Gore JC. In vivo measurement of adc change due to intravascular susceptibility variation. Magn Reson Med 1999;41:236–24
Stejskal EO, Tanner JE. Spin diffusion measurements: spin echoes in the presence of a time-dependent field gradient. J Chem Phys 1965;42:288–292

Main menu

User menu

Search

American Journal of Neuroradiology

References

References

News and Updates

Resources

Opportunities

American Society of Neuroradiology