4D CT Angiography More Closely Defines Intracranial Thrombus Burden Than Single-Phase CT Angiography

Assessing Clot Burden and Thrombus Length

The present study shows that tMIP reconstructions from 4D-CTA more closely outline intracranial clot burden than conventional arterial-phase spCTA. Although the appearance of clot burden can be identical on spCTA and tMIP, the latter revealed significant additional information on clot extent in approximately one-third of our population. The distal thrombus end was identified more often on tMIP than on spCTA (both raters identified it in 72% versus 45% of cases). Because tMIP can be reconstructed from volumetric CTP examinations without additional contrast or radiation exposure, the presented approach can be easily integrated into existing CT stroke imaging protocols.

Importantly, it is conceivable that thrombi could also be more closely approximated by adjusting the speed and triggering delay of spCTA to reveal delayed collateral flow. However, intracranial bolus passage is difficult to predict, and delaying the scan too much may dramatically reduce arterial contrast. Adjustment of spCTA timing thus does not seem justified on the basis of our present results, especially considering that tMIP can be obtained simply by adding another reconstruction, without otherwise altering the multi-modal CT protocol.

The overall observed difference in CBS between tMIP and spCTA was rather small (1 point difference of the median), in part because of patients with identical CBS on tMIP and spCTA (n = 22). When a difference was present, a commonly observed pattern was that tMIP showed patency of M2 branches not seen on spCTA. The difference in CBS is best explained by the increased sensitivity of tMIP for delayed collateral flow reaching the vasculature distal to the occlusion site. Indeed, in the initial description of the CBS it was pointed out that the score depends on delayed collateral flow.¹ The authors conclude that this may actually be a strength of the score, because decreased collateral flow (which in turn will decrease the CBS on spCTA) has been associated with larger infarcts and poorer outcome. The tMIP technique may thus weaken the prognostic power of the CBS because temporal fusion makes strong, early collateral flow virtually indistinguishable from delayed, weaker collaterals, leading to identical CBS despite different physiologic implications. Interestingly, we observed a (nonsignificant) trend toward a higher presenting NIHSS with lower CBS on spCTA; this correlation was less apparent on tMIP. Thus, we deem it possible that diminished collateral flow (leading to lower CBS scores on spCTA but not necessarily on tMIP) is accompanied by more severe clinical deficits on presentation.

To validate both spCTA and tMIP against a reference, we assessed clot length in patients with M1 occlusion who also had an unequivocal HMCAS and sufficient collateral flow according to at least 1 rater. Our results show an excellent correlation between the length of the filling defect on tMIP and the length of the HMCAS, confirming that tMIP may adequately outline middle cerebral artery thrombi. Correlation of HMCAS with spCTA was not significant, probably because of inaccuracy in estimating the filling defect on spCTA, depending on collateral flow. Previous investigators have shown a strong correlation of the HMCAS and the spCTA filling defect, but they specifically selected patients with sufficient collateral flow and a filling defect matching the HMCAS,¹⁴ which may explain this discrepancy. In 2 patients with long (>20 mm) clots, the tMIP filling defect was measured >5 mm shorter than the HMCAS, which we tend to attribute to the slightly different measurement techniques used (curved analysis for HMCAS, orthogonal lines for filling defects).

Thin-section NCCT has previously been used elegantly to assess clot length (measured as HMCAS length) and predict response to intravenous thrombolysis.⁵ Compared with tMIP, this represents a technically less demanding approach to determine middle cerebral artery clot length. However, the technique may be affected by clot composition, because erythrocyte-rich thrombi have been shown to display higher attenuation values than platelet-rich thrombi.^6,8 Furthermore, the presence of vascular calcifications or a high hematocrit may mimic the HMCAS and thus decrease specificity.¹⁴ Thus, tMIP could be helpful in quantifying thrombus length when an HMCAS is not unequivocally seen or when the longitudinal extent of the HMCAS is uncertain. In addition, it is conceivable that an erythrocyte-rich, high-attenuation, and hence well-visible thrombus may be accompanied by appositional, platelet-rich, low-attenuation thrombi, which would lead to discrepancies between the HMCAS and CTA imaging. This could hypothetically account for some of the presently observed differences between tMIP and HMCAS; however, this did not affect overall correlation of tMIP and HMCAS in our comparably small sample size. Particularly in patients with weak collateral flow, any arterial segment not opacified on tMIP could theoretically harbor thrombi or stagnant blood, potentially in different stages of the clotting process. Correlation with thin-section NCCT could then reveal which parts correspond to high-attenuation thrombus. For these reasons, we propose that CT angiographic techniques may in some cases offer complementary information on clot physiology compared with the assessment of the HMCAS. The value of tMIP, then, lies in the ability to depict the complete extent of vessels perfused within the acquisition time.

One of our study's limitations is that thin-section NCCT was not routinely obtained. Instead, we used temporal average thin-section reconstructions from the early, nonenhanced phase of the CTP examination to assess the HMCAS, which may have affected sensitivity and comparability of our results. We observed an unequivocal HMCAS in 33% of patients with M1 occlusion, which is low compared with recently reported sensitivity and specificity values of the HMCAS on thin-section NCCT.¹⁵ Four-dimensional CTA was reconstructed with a higher section width and increment than spCTA (1.5/1.0 mm versus 0.75/0.4 mm), which was chosen as a compromise between a manageable number of images and acceptable image quality. Other limitations include the presence of more venous enhancement on tMIP than on spCTA, which may in some cases make evaluation of the arterial tree more difficult. However, this was not a limiting factor for arterial analysis in our patients. Slight venous contamination was also invariably present on spCTA images because of residual contrast material from the immediately preceding CTP examination. We also cannot exclude that additional vessel segments would become visible on tMIP if the acquisition time of the CTP examination would be increased beyond the presently used 45 seconds. Finally, the retrospective design of our study carries the risk of selection bias, particularly because we only assessed patients undergoing endovascular treatment. However, this allowed us to assess early recanalization on conventional angiography after bridging thrombolysis. We did not observe a clear relation between early recanalization and clot burden, which we tend to attribute to the low sample size of early recanalizers (n = 6). This low early recanalization rate (11%) emphasizes previous reports on this matter.⁴ It is important to note that recanalization was assessed relatively early in our study: Thrombolysis was initiated as soon as possible after interpretation of NCCT; the mean 88 minutes between CT and DSA can therefore be used to approximate the maximum duration between thrombolysis and assessment of early recanalization. This may have contributed to the relatively low recanalization rate observed. Assessing tMIP and other 4D-CTA reconstructions in predicting response to standard intravenous therapy seems desirable; however, this would require a different patient collective to compare outcomes after intravenous therapy only.

Clinical Implications

Accurate information on the extent of thrombus is important for adequate treatment planning. Previous studies have consistently shown that larger, more proximal clots often do not respond to intravenous thrombolysis and are associated with worse clinical outcome.^{1⇓⇓⇓–5} As a result, more aggressive endovascular recanalization procedures are often considered in these patients. With an increasing number of available neurovascular medical and device-assisted treatment options, precise definition of the target lesion, that is, the occluding thrombus, is desirable. Animal model evidence has shown that longer clots (>10 mm) are associated with decreased procedural success and increased rates of complications such as distal embolization during mechanical thrombectomy.¹⁶ Evaluation of the patency of M2 branches may also be particularly helpful when mechanical thrombectomy is considered, further emphasizing the need for precise thrombus delineation from the interventionalist's perspective. With the recent halting of the Interventional Management of Stroke III trial, it is becoming increasingly clear that adequate selection of patients likely to benefit from a specific type of endovascular therapy is crucial in establishing evidence of clinical efficacy. Besides clinical variables and imaging strategies directed at defining salvageable brain tissue, clot characteristics such as length,⁵ location,^17,18 composition,^6⇓–8 and degree of luminal occlusion^19⇓–21 should be considered as potential predictors of the clinical efficacy of different available medical and endovascular treatment modalities.