CT perfusion predicts secondary cerebral infarction after aneurysmal subarachnoid hemorrhage

After aneurysmal subarachnoid hemorrhage (SAH), the noninvasive detection of vasospasm and the prediction of secondary cerebral infarction (SCI) may help to indicate invasive angiographic diagnosis and therapy of vasospasm. In clinical practice this is hampered by two reasons: First, the clinical examination may be difficult to interpret especially in these patients and second, transcranial Doppler sonography (TCD) as the routine noninvasive modality is not sensitive for the detection of angiographic vasospasm¹ and is only weakly associated with SCI as one important related complication after SAH.² CT perfusion (CTP) is increasingly being applied for the assessment of cerebrovascular disorders. In this study, we performed native CT, CTP, and TCD at narrow intervals after SAH to investigate the predictability and time of prediction for later infarction on native CT.

METHODS

Fifty consecutive patients presenting with aneurysmal SAH were primarily eligible for the study, which was approved by the local ethics committee with informed consent of the patients or their relatives. Thirty-eight patients were part of the final data analysis after exclusion of 12 patients for the following reasons: early death (n = 5), nonconformity with study protocol (n = 4), motion artifacts (n = 2), history of severe allergic reaction (n = 1). On admission, patients underwent a standardized neurologic examination. Hunt and Hess grade and neurologic deficits were documented. The study protocol scheduled daily neurologic examinations and TCD over 2 weeks, if possible by the same investigator. Combined native CT and CTP were conducted at regular intervals (3 ± 2, 5 ± 2, 7 ± 2, 14 ± 2 days after admission). An average of 3.5 CT/CTP and 10.7 TCD examinations were performed per patient. The target condition was SCI defined as complete late infarction on CT between 3 and 14 days after SAH. SCI developed in n = 14 patients (n = 24 without SCI). If complete infarction was already present on the first postprocedural native CT (i.e., before day 3 after hemorrhage), it was considered therapy related and excluded. For the determination of irreversible brain infarction on native CT, common diagnostic criteria were applied.³ For the prediction of the target condition, only CTP and TCD examination dates before manifest infarction on native CT were included. Control data were the CTP and TCD examinations of the patients not developing SCI at comparable time points after admission.

For native CT slice thickness was 5 mm for the posterior fossa and 8 mm for the above cerebrum (Somatom Plus 4 Volume Zoom, Siemens, Erlangen, Germany). For CTP, two adjacent 10-mm slices were positioned at the level of the basal ganglia with the same angulation as for the native CT. A bolus of 50 mL of nonionic contrast medium (Imeron 400, Bracco, Konstanz, Germany) was administered by a power injector into a central venous catheter at a flow rate of 4 mL/s followed by 30 mL of saline. Four seconds after beginning of the bolus, 40 images were collected at each slice level at a rate of two images per second (120 kV, 110 mAs, matrix 512 × 512). The effective dose of a single CTP scan for these parameters was previously estimated to be 5.5 mSv (Application Guide Somatom Plus 4 Volume Zoom, Siemens AG). Slice positioning was always above the orbital roof in the supraorbitomeatal direction to protect the lenses, corresponding to a mean lens dose of less than 30 mGy per study. For CTP analysis a commercially available software was used (Perfusion CT, Siemens). CTP color maps were qualitatively assessed by two independent observers blinded to patient data. A positive visual assessment was noted for side-to-side asymmetries suggesting a decrease in cerebral blood flow (CBF), cerebral blood volume (CBV), or an increase in time to peak (TTP). Interobserver agreement was reached in 98% of the color-map judgments (Cohen κ = 0.86, p < 0.0001). Quantitative CTP measures were gained from six regions of interest (ROIs) defined a priori. These ROIs were manually outlined by one investigator to comprise the cortical gray matter territory of the anterior cerebral artery (ACA), middle cerebral artery (MCA), and posterior cerebral artery (PCA) of each hemisphere. To balance limitations inherent to any model assumption underlying the calculation of CBF and CBV, and especially to correct for TTP delays rather related to low cardiac output, quantitative values were normalized from side to side. TCD was performed daily using a Companion TCD system (EME, Überlingen, Germany). As in previous investigation, for qualitative analysis the mean flow velocities in the ACA, MCA, and PCA territories were dichotomized at 120 cm/s to suggest vasospasm possibly leading to SCI.⁴

Diagnostic accuracy of qualitative CTP and TCD assessments was determined by sensitivity, specificity, and negative and positive predictive values (NPV, PPV) with corresponding 95% CIs. The risk of developing SCI as a function of quantitative measures was analyzed with binary logistic regression, thereby setting no threshold value for the target condition.

RESULTS

Patient data and qualitative CTP and TCD ratings.

Mean patient age was 50.4 years (range 34 to 82 years). Distribution of Hunt and Hess grades on admission was n_I = 9, n_II = 10, n_III = 15, and n_IV = 4. The median Fisher grade was 4 in both groups (range 3 to 4). The median interval between the predictive CTP session before complete SCI on native CT was 3 days (range 2 to 5 days). TTP color maps predicted SCI with 0.93 sensitivity (CI: 0.7 to 1.0), 0.67 specificity (CI: 0.53 to 0.7), 0.94 NPV (CI: 0.75 to 1.0), and 0.62 PPV (CI: 0.46 to 0.66). For CBF sensitivity was 0.29 (CI: 0.11 to 0.35), specificity 0.96 (CI: 0.86 to 1.0), NPV 0.7 (CI: 0.62 to 0.73), and PPV 0.67 (CI: 0.8 to 0.99). CBV showed 0.21 sensitivity (CI: 0.06 to 0.28), 0.96 specificity (CI: 0.87 to 1.00), 0.68 NPV (CI: 0.62 to 0.7), and 0.75 PPV (CI: 0.23 to 0.99). Any combination of CTP measures did not improve diagnostic performance.

Daily TCD before manifestation predicted SCI with 0.58 sensitivity (CI: 0.32 to 0.82), 0.5 specificity (CI: 0.37 to 0.62), 0.71 NPV (CI: 0.52 to 0.87), and 0.37 PPV (CI: 0.2 to 0.52). All TCD examinations over 14 days of observation detected SCI with 0.77 sensitivity (CI: 0.53 to 0.93), 0.38 specificity (CI: 0.24 to 0.46), 0.75 NPV (CI: 0.49 to 0.93), and 0.4 PPV (CI: 0.27 to 0.49). Figure 1 displays the sensitivity plot of qualitative CTP and TCD ratings in the time course before and after SCI.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Prediction of secondary cerebral infarction

Sensitivity plots of transcranial Doppler (TCD) and the time-to-peak (TTP) parameter of CT perfusion (CTP) for the prediction of secondary cerebral infarction. Different time points before and after manifest infarction on native CT are displayed. Note the marked difference in sensitivity for the two noninvasive modalities early before infarction.

New delayed neurologic deficits could be clearly localized to a vascular territory and had an exact time of onset in five patients. In these patients territorial TTP findings corresponded well to the clinical symptoms. All other cases of new delayed clinical deterioration were either nonlocalizing or the clinical findings were difficult to interpret, which was often related to sedative measures. Reangiography was performed in six patients. The CTP results of the patients with reangiography were consistent with the overall analysis predominantly showing a marked TTP delay in territories affected by moderate or severe angiographic vasospasm. Figure 2 displays CTP color maps and native CT images for one representative patient before and after SCI.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Native CT and CT perfusion color maps of a 60-year-old patient 2 days after rupture of a right internal carotid artery aneurysm that was clipped

On native CT no parenchymal hypodensity indicating early infarction is present. The time-to-peak (TTP) map demonstrates a marked delay in the anterior and posterior portion of the middle cerebral artery (MCA) territory. To a lesser degree also, cerebral blood flow (CBF) and cerebral blood volume (CBV) maps reveal regional hemodynamic impairment. On following native CT 4 days later, partially hemorrhagic cerebral infarction is present in the anterior and posterior portion of the MCA territory. SAH = subarachnoid hemorrhage.

DISCUSSION

Our results demonstrate high sensitivity and considerable specificity of CTP for the early prediction of SCI. A recent study on CTP after aneurysmal SAH for the first time established a clear relation between perfusion parameters on admission and subsequent delayed cerebral ischemia.⁵ With this investigation, we add estimates from observations at regular intervals early after admission and document the time between the predictive diagnostic tests and SCI as completed delayed infarction on native CT. With predominant sensitivity TTP was the best-performing CTP parameter. Obviously, TTP is inherently sensitive to a delay in tissue perfusion as occurs early in vasospasm, which is the most relevant cause for SCI.⁶ On this account, the diagnostic value of TTP is especially promising for the prediction of SCI and the early detection of potentially underlying vasospasm. It is therefore different from the value of CBF and CBV, which qualified to accurately indicate the critical decrease of tissue perfusion seen during acute cerebral ischemia.⁷

As an important additional finding, the visual assessment of CTP color maps as performed in the clinical routine was not inferior to quantitative ROI analysis based on vascular territories defined a priori. Such arbitrary stipulation of ROIs for quantitative analysis and a lack of validity of perfusion measures may explain best why there was no substantial diagnostic gain over visual assessment. To some extent this limitation of absolute measures is compensated by intraindividual side-to-side normalization, which, however, complicates the analysis in any case of bilateral infarctions regularly seen if vasospasm is diffuse. It is also important to acknowledge that the time window eligible for therapeutic intervention is not necessarily as wide as observed and that this study cannot prove effective treatment of potentially underlying vasospasm or prevention of secondary infarction during this interval. In addition, at the time of the predictive session, secondary infarctions in the hyperacute stage or those too small to present as demarcated hypodensity may have been present but eluded detection on native CT.

The association of TCD measures with SCI was low. This is fairly consistent with earlier findings suggesting that TCD, although specific for the detection of angiographic vasospasm, is not sensitive¹ and only weakly correlated with ensuing late infarctions.^2,8,9

The impact of improved diagnostic management by CTP on a defined and well-documented clinical endpoint has yet to be demonstrated. However, as suggested by earlier reports,¹⁰ our results help to further specify the usefulness of this widely available technology in that secondary infarctions as one important complication after SAH are predictable with high sensitivity. A further increase in its diagnostic performance may be achieved in the future through the coverage of wider anatomic regions and improved validity of quantitative measures. At present already, even though CTP examines only confined regions, especially the employment of CTP during early observation seems reasonable and is supported by this study. During this period, particularly the patients with clinical findings difficult to interpret and inconclusive TCD will benefit in that the impending risk for later infarction may be better assessed. The very sensitive and early prediction of SCI by CTP may improve the timely indication of invasive angiography and prompt early therapy of potentially underlying vasospasm.

ACKNOWLEDGMENT

E. Klotz (Siemens AG, Medical Solutions, Forchheim, Germany) assisted us with excellent technical advice in the selection of the CTP protocol and data postprocessing. He also kindly provided measures of radiation exposure for the employed CT unit. The authors thank Prof. G. Stoll (Department of Neurology, University of Würzburg) for critical reading of the manuscript and valuable comments. Dr. H.P. Schlake especially contributed to data collection from the neurosurgical intensive care unit. M. Pham and A.J. Bartsch were supported by Dr. D. Leising from the Center of Clinical Studies of the University of Würzburg in data management and statistical analysis.