METHODS

Between 1989 and 2000, 162 consecutive patients with acute embolic stroke involving the anterior circulation were admitted to St Mary’s hospital, Kurume, Japan, and underwent cranial CT within six hours of stroke onset. Embolic stroke was defined on the basis of the criteria for cardiogenic brain embolism established by the Cerebral Embolism Task Force.¹³ Fifty four patients who did not undergo cerebral angiography within two hours of their CT examination were excluded, as were three patients with isolated infarction in the territory of the anterior cerebral artery. Thus the final cohort included 105 patients.

Even in patients treated with intravenous thrombolysis, we generally carried out cerebral angiography. A cardiac or transcardiac embolic source was present in 86 patients. The remaining 19 had aortic or carotid arterial atheroma. Neurological deficits on admission were evaluated using the National Institutes of Health stroke scale (NIHSS).

Non-contrast cranial CT was done with a slice thickness of 10 mm, looking for infarcts or early CT signs (LIR, ALN, HSE, and HMCAS), according to definitions used in Moulin’s previous study (fig 1).⁸ Briefly, LIR was defined as decreased precision in delineating the grey-white interface at the lateral margin of the insula; ALN was defined as a decrease in density involving the lentiform nucleus area, including loss of precise delineation of the area; HSE was defined as decreased contrast density inducing a loss of precise delineation of the grey-white interface at the margins of the cortical sulci, corresponding to a localised mass effect; and HMCAS was defined as spontaneous high contrast in the middle cerebral artery. The findings were agreed between two stroke neurologists blinded to the patients’ clinical features and to the results of the arteriogram. Both were familiar with neuroradiology.

Conventional angiography (n = 46) or digital subtraction angiography (n = 59) was done using transfemoral catheterisation. We divided the patients into three groups according to the arterial occlusion site: internal carotid artery occlusion; occlusion of the sphenoidal portion of middle cerebral artery (M1 occlusion); and occlusion of the middle cerebral artery at the bifurcation or the peripheral branch (branch occlusion).

Thrombolytic treatment was given to some patients within six hours of stroke onset.

We evaluated the following three indicators of therapeutic outcome in the patients: complete or partial recanalisation of the occluded artery documented on angiography one hour after thrombolysis; symptomatic or asymptomatic haemorrhagic transformation of the infarct documented on CT during the hospital admission; and modified Rankin scale score four weeks after stroke onset.

Statistics

Data are presented as mean (SD). In the univariate analysis, categorical variables were compared by the χ² test, continuous variables by an unpaired Student t test, and scoring variables by the Mann–Whitney U test. Multivariate analysis was done by logistic regression to identify the predictor variables for each arterial occlusion site. Differences were considered significant at a probability (p) value of < 0.05.

RESULTS

Sixty two men and 43 women aged 66 (10) years (range 42 to 80) were studied. Median NIHSS score on admission was 16 (range 3 to 30).

CT was undertaken at a mean of 112 (63) minutes (range 13 to 357) after stroke onset. ALN, LIR, or HSE, each indicating parenchymatous ischaemia, was present in 65 patients (62%). The prevalence of each of these three signs and their combinations is shown in table 1. HMCAS was present in 27 patients (26%), 23 of whom also had early parenchymatous signs. A fresh brain infarct was present in two patients (2%) and old infarcts in 32 (30.5%).

Angiography was done at a mean of 210 (68) minutes (range 55 to 410) after stroke onset (99 (48) minutes after CT). Internal carotid artery occlusion was present in 29 patients (28%), M1 occlusion in 34 (32%), and branch occlusion in 38 (36%).

LIR was the most frequent sign in patients with internal carotid artery occlusions (sensitivity 55%), ALN in those with M1 occlusions (sensitivity 65%), and HSE in those with branch occlusions (sensitivity 47%). Among combinations of these three signs, LIR+ALN+HSE seemed to indicate internal carotid artery occlusion (34%) or M1 occlusion (24%), LIR+ALN to indicate M1 occlusion (24%), and isolated HSE without LIR or ALN to indicate branch occlusion (37%). HMCAS seemed to indicate internal carotid artery occlusion (34%) or M1 occlusion (32%).

Possible predictive factors for arterial occlusion site were identified by univariate analysis for age, sex, NIHSS score, and the presence of each of the early signs alone or in combination (table 2). After multivariate analysis using the factors as independent variables, LIR was related independently to internal carotid artery occlusion (odds ratio (OR), 2.8 (95% confidence interval (CI), 1.2 to 6.8)); NIHSS score ⩾ 16 (OR 3.4 (1.2 to 9.4)) and ALN (OR 2.9 (1.2 to 7.4)) to M1 occlusion; and NIHSS score < 16 (OR 5.0 (1.6 to 10.0)) and isolated HSE (OR 12.8 (3.2 to 51.5)) to branch occlusion. HMCAS, either alone or in combination with other signs, was not an independent indicator of any arterial occlusion site.

Thrombolytic treatment during the hyperacute stage of stroke was given in 81 patients, 52 intra-arterially and 29 intravenously; 25 were given tissue plasminogen activator (tPA) and 56 were given urokinase. Among these treated patients, 34 (42%) showed recanalisation of the occluded artery, 44 (54%) showed haemorrhagic infarction, and 29 (36%) showed good functional prognosis corresponding to a modified Rankin scale score of between 0 and 2. Arterial recanalisation was documented less often in patients with than without LIR (p < 0.03, table 3). Haemorrhagic infarction was more often found in patients with LIR (p < 0.03), ALN (p < 0.02), HMCAS (p < 0.05), or combinations of these signs (p < 0.02) than in patients without those signs. None of the CT signs was significantly correlated with good functional prognosis. The site of arterial occlusion in these 81 patients was internal carotid artery in 18 (62% of all patients with internal carotid artery occlusion), M1 in 32 (94%), and the middle cerebral artery branch in 31 (82%). Arterial recanalisation was documented in 22% of patients with internal carotid artery occlusion, in 44% of those with M1 occlusion, and in 52% of those with branch occlusion (p > 0.1); haemorrhagic infarction occurred in 72%, 69%, and 29%, respectively (p < 0.001); a good functional outcome was likely in 11%, 41%, and 45%, respectively (p < 0.05).

DISCUSSION

This is the first study on early CT signs as predictors of the arterial occlusion site in embolic stroke using a relatively large population. All parenchymatous signs and their combinations were useful for predicting internal carotid artery, M1, or branch occlusions. Our results assessing the vascular lesions causing infarction make a good contrast with those of Moulin et al,⁸ who assessed subsequent infarcts.

Loss of insular ribbon

The strong correlation between LIR and internal carotid artery occlusion in our present study is not in agreement with the suggestion in the original study³ that LIR might result from middle cerebral artery occlusion distal to the lateral striate arteries. In contrast, the dominant appearance of extended infarcts in the middle cerebral artery territory in patients with LIR in Moulin’s later study⁸ appears to support our results to some extent. LIR hardly ever appeared alone in the present study, and more than half the patients with LIR also had ALN and HSE. The concomitant presence of these three signs also seemed to indicate internal carotid artery occlusion. After the European cooperative acute stroke study (ECASS) trial first indicated that hypoattenuation in more than 33% of the middle cerebral artery territory was a predictor of fatal outcome with thrombolytic treatment,¹⁴ subsequent studies stressed this major early finding and paid little attention to LIR.^2,15,16 In contrast, the rt-PA stroke study group of the National Institute of Neurological Disorders and Stroke (NINDS) indicated that patients treated with tPA had a favourable outcome even if they had hypoattenuation in more than 33% of the middle cerebral artery territory.¹⁷ Thus the diagnostic significance of this major early finding in relation to the use of thrombolysis is in dispute. Based on the present results, a combination of LIR and > 33% hypoattenuation of the middle cerebral artery territory may be a promising warning sign against the use of thrombolysis.

Attenuation of the lentiform nucleus

In the original studies,^4,5 all the acute stroke patients with angiographically proven internal carotid artery or M1 occlusion had ALN, presumably because the lentiform nucleus is fed by lenticulostriate arteries from M1 without collateral flow from cortical anastomoses. In the present study, ALN was also highly predictive of M1 occlusion, but was not present in all patients with this type of lesion. This might reflect the timing of CT. The presence of ALN in 16% of patients with branch occlusions appeared to result from variation in the lenticulostriate arteries, which arise from the middle cerebral artery branch in around 20% of cases.¹⁸

Hemispherical sulcus effacement

In contrast to LIR or ALN, HSE reflects cortical ischemia. Isolated HSE without LIR or ALN was highly indicative of branch occlusion in this study, and of a partial superficial infarct in the middle cerebral artery territory in the previous study.⁸ Thus isolated HSE in stroke patients seems to be a good indication for intravenous thrombolysis.¹²

Treatment

Thrombolytic treatment was selected on the basis of the angiographic findings in this study. Many patients with internal carotid artery occlusion did not undergo thrombolysis. In spite of this selection bias, there were some interesting correlations between early CT signs and therapeutic outcome. Patients with LIR had poor arterial recanalisation, presumably because patients with internal carotid artery occlusion have a poor recanalisation rate and LIR has a strong correlation with internal carotid artery occlusion. Patients with HSE did not often have haemorrhagic infarction, presumably because patients with branch occlusion also did not often have haemorrhagic infarction, and HSE was strongly correlated with branch occlusion. The relatively small patient population related to the above mentioned selection bias may be the reason why early CT signs did not predict functional outcome in the chronic stage.

Limitations

Patients with high NIHSS scores on admission often failed to get their angiography within two hours after CT and so were excluded from the study. Internal carotid artery occlusion might well be the leading vascular lesion in these excluded patients, because this often causes large and fatal infarction in the middle cerebral artery territory.¹⁹ Thus the exclusion of such patients may have affected our results with respect to internal carotid artery occlusion. The strong correlation of a high NIHSS score with M1 occlusion, but not with internal carotid artery occlusion, seems be a reflection of this possible bias. Another limitation of our results was that the sensitivity of each early ischaemic sign is rather low. In practical clinical situations, combining the information obtained from CT with, for example, carotid ultrasonography seems to be important for decisions about treatment in acute stroke. The CT analysts in this study were stroke specialists who were familiar with neuroradiology. However, even for experienced stroke specialists, there is considerable lack of agreement in recognising and qualifying the early CT changes.²⁰ Improvements in diagnostic methods for identifying early CT signs are expected. Finally, we should bear in mind that CT equipment upgrades over the 11 year period of the study may have introduced some differences in image quality that could have influenced our results.

Conclusions

A combination of early parenchymatous CT signs is important in predicting the arterial occlusion site in the acute stage of ischaemic stroke.