Avoiding “Pseudo-Reversibility” of CT-CBV Infarct Core Lesions in Acute Stroke Patients After Thrombolytic Therapy
The Need for Algorithmically “Delay-Corrected” CT Perfusion Map Postprocessing Software
Background and Purpose—Rarely, acute ischemic stroke (AIS) patients have pretreatment CT-CBV abnormalities larger than final infarct volumes. We sought to determine: (1) the prevalence of CT-CBV “reversibility” in AIS patients treated with thrombolytic therapy, and (2) whether the presumed tissue salvage of these CT-CBV lesions depends on the CTP software.
Methods—We reviewed the admission CT-CBV maps (calculated with an algorithm sensitive to tracer arrival time) and follow-up images of 148 AIS patients who received thrombolytic therapy. When the follow-up infarct appeared smaller than the admission CT-CBV lesion, the CTP source images were reprocessed using “delay-correction” software (GE, CTP 4). Original and “delay-corrected” CT-CBV ischemic lesion volumes were compared to each other and follow-up infarct volumes using the Student t test.
Results—11/148 (7.4%) patients had admission CT-CBV larger than follow-up lesions (mean difference −69.5 cc, range −146.0 to −14.0 cc; P<0.05). For all patients, the admission CT-CBV lesions were smaller on the delay- versus nondelay-corrected maps (mean difference −83.1, range −233 to −2 cc; P<0.05). Only 2 patients had delay-corrected CT-CBV lesions larger than follow-up infarctions, with a 12- to 17-cc difference in volume. 7/9 of the remaining patients had extracranial hemodynamic factors potentially delaying tracer arrival, including atrial fibrillation (AF; n=4), congestive heart failure (CHF; n=4), or extracranial internal carotid artery (ICA) stenosis (n=1).
Conclusion—True “reversibility” of CT-CBV “core” lesions in AIS patients after thrombolytic therapy is rare, with small volumes of “salvaged” tissue. Pseudoreversibility of core lesions in standard CT-CBV maps can be avoided by using specific algorithmically optimized delay-correction software. Further investigation is warranted to determine whether this finding applies to algorithms provided by other vendors.
CT-CBV maps are constructed by some commercial software packages using deconvolution with singular value decomposition (SVD) algorithms. CT-CBV hypovolemic lesions, similar to diffusion weighted imaging (DWI) lesions, are thought to reflect infarct core in acute ischemic stroke (AIS) patients.1 However, we have encountered cases where, after thrombolytic therapy, follow-up infarct volumes are smaller than admission CT-CBV lesions.
We speculate that “reversibility” of CT-CBV lesions might: (1) result from early reperfusion, or (2) be overestimated when the modeling assumptions underlying the CT-perfusion (CTP) algorithm are not met because SVD is sensitive to contrast bolus delay.2 Specifically, we hypothesized that CTP algorithms that correct for delay might reduce the occurrence of CT-CBV overestimation, especially in patients with severe extracranial hemodynamic compromise. Therefore, we sought to: (1) determine the prevalence of CT-CBV “reversibility” in AIS patients treated with thrombolytic therapy, and (2) determine whether the presumed tissue salvage of these CT-CBV lesions is dependent on the CTP software used.
Patients and Methods
This HIPAA compliant study was approved by our Institutional Review Board. We reviewed the records of patients admitted to Massachusetts General Hospital with AIS treated with intravenous (IV) or intraarterial (IA) thrombolytics from March 2003 to August 2007. Patients with (1) anterior circulation infarctions, (2) admission CTP, and (3) follow-up CT/MR within 6 months were included.
CTP was performed on a multi-detector scanner (Light Speed, GE Medical Systems), as a 60-second cine series, beginning 5 seconds after a 40-mL injection of Isovue 370 (Bracco) at 7 mL/s. Parameters were 80kVp, 200mA, 1-second rotation, FOV-22 cm, matrix 512×512. On the 16-slice scanner, 4 5-mm slices covered 2 cm. On the 64-slice scanner, 8 5-mm slices covered 4 cm.
CTP maps were created using 2 deconvolution software versions (CT-Perfusion 3(CTP3) and the most current version of CTP4 (GE Medical Systems). CTP3 deconvolves the tissue contrast function with the arterial input function (AIF) using SVD, sensitive to contrast arrival delay. CTP4 is algorithmically optimized and compensates for delay by shifting the tissue curve baseline to synchronize it with the AIF curve. The AIF was the contralateral ACA. The venous function was the superior sagittal sinus.
Unenhanced CT was performed on a multi-detector scanner (Light Speed; GE Medical Systems) with 5-mm axial images, 140-kVp, 170-mA, 1-second rotation.
Axial FLAIR images were obtained on a 1.5-Tesla scanner (Signa; GE Medical Systems), with TR/TE/TI of 10002/141/2200-ms; FOV-24 cm; matrix 256×256; thickness 5-mm, 1-mm gap; 1 average.
An experienced neuroradiologist visually compared all CT-CBV (calculated with CTP3) with follow-up lesions and reviewed reports comparing follow-up infarct volumes to initial CT-CBV. When both reports and visual inspection suggested that follow-up lesions were smaller than admission CT-CBV lesions, CT-CBV maps were reprocessed using CTP4. When reports and visual inspection were discrepant, a second neuroradiologist, in consensus with the first, visually compared the CT-CBV with follow-up lesions; if the consensus reading determined that the follow-up infarction was smaller than the initial CT-CBV lesion, CT-CBV maps were reprocessed using CTP4. Two experienced technologists with advanced training in acquisition and processing of CTP data manually segmented all CT-CBV and follow-up images to calculate volumes with a commercially available software program (Analyze 7.0, Analyze Direct). All outlines were checked and adjusted as necessary by an experienced neuroradiologist. Each segmentation was performed in a randomized fashion without knowledge of clinical data or of the results of any other segmentation. By visual inspection, we defined “adequate” arterial, venous, and tissue curves as those with an identifiable peak and full wash-in and wash-out of contrast without truncation. The algorithm chooses the curves by sampling adjacent pixels in an ROI to identify the peak maximal enhancement.
The CTP3, CTP4, and follow-up lesion volumes were compared with a Student t test. P<0.05 was significant.
The Table presents patient characteristics. On CT-CBV maps processed with CTP3, 11/148(7.4%) patients had lesions that were larger than follow-up infarctions (Figure 1); mean volume difference was −69.5 cc (range −146 to −19 cc; P<0.005). 7/ 71(10%) patients who received IAT and 4/77(5.2%) patients who received IVT had such lesions.
CT-CBV maps were not processed with CTP4 for 2 patients: In 1 patient (No. 9) with CHF, the AIF curve did not reach its peak before the CTP acquisition ended. Another patient (No. 10) had complete ipsilateral ICA occlusion with tissue curve truncation. Lesions were smaller on CT-CBV maps processed with CTP4 versus CTP3 in the remaining 9 patients (mean difference −83.1, range −233 to −2 cc; P<0.05) who had adequate arterial, venous, and tissue curves.
On CT-CBV maps processed with CTP4, only 2 patients (Nos. 3 and 5; Figure 2) had lesions that were larger than follow-up infarctions, with 12 to 17 cc difference in volume. Both had M1 occlusions with recanalization after endovascular treatment. These were consistent with true CBV reversal. The remaining 7 patients had delayed bolus arrival in tissue curves compared to the contralateral side. Five had possible contributing extracranial factors, including AF (n=3), CHF (n=3), or ipsilateral extracranial ICA stenosis (n=1; Table).
We demonstrated that standard CT-CBV maps calculated with deconvolution can overestimate infarct core, but when CT-CBV maps are processed with algorithmically optimized “delay-correction” software, recovery of tissue with a low CBV is rare. Specifically, standard CT-CBV maps demonstrated abnormalities larger than final infarct volumes in 11/148 (7.4%) AIS patients treated with thrombolytic therapy. However, only 2 patients had “delay-corrected” CT-CBV abnormalities that were larger than final infarct volumes, and the decreases in lesion size were small. These findings correlate with reports demonstrating that final infarcts are rarely smaller than initial DWI abnormalities.3
Delayed tracer bolus arrival can be attributed to extracranial and intracranial factors. Extracranial factors include AF, CHF, and ICA stenosis. Intracranial causes include thrombus extent and poor collateralization. In patients with these factors who demonstrate a large CT-CBV defect, arterial, venous, and tissue curves must be inspected. There is expert consensus that CTP acquisition should last for at least 45 seconds to avoid CBV-time density curve truncation and CBV lesion overestimation.4 Special attention must be paid to delay between arterial and tissue curves, and delay-correction algorithms should be applied to accurately identify infarct core.
It is important to note that only quantification of CT-CBF has been validated.5 We are aware of no study that has validated the accuracy of quantitative CT-CBV measurements. Furthermore, for the perfusion algorithm described herein, the calculated CT-CBV is dependent on the impulse residue function produced by SVD; deconvolution is felt necessary to avoid errors from recirculation.5,6 Prior MR studies have shown that CBV is unaffected by tissue contrast arrival delay.2 MR-CBV, based on the central volume theorem, is the integral under the tissue contrast curve, and unaffected by delay.
By reducing underestimation of cerebral perfusion and overestimation of ischemic tissue, software with delay correction may improve identification of infarct core. This could improve our ability to differentiate irreversibly infarcted from salvageable tissue and could improve stroke patient treatment. Although using 2 different CT scanners is a potential limitation, the acquisition parameters and cine mode are identical for both scanners and the CTP algorithms should process the data similarly. Furthermore, we would like to stress that although our results are promising, algorithmically optimized software with delay correction should be verified in a large number of stroke centers and the standardization of adequate CTP algorithms is crucial for management of acute ischemic stroke patients. In addition, although MR-CBV calculation is not affected by delay, MR-CBF and MR mean transit time parameters are affected by tissue contrast arrival delay and optimal MR perfusion algorithms with delay correction are also needed for assessment of acute ischemic stroke patients.
True “reversibility” of CT-CBV “core” lesions in AIS patients after thrombolytic therapy is rare. Pseudoreversibility of core lesions can be avoided by using specific delay-correction software. Further investigation is warranted to determine whether this finding applies to algorithms provided by other vendors.
Michael H. Lev: GE Healthcare (modest), CoAxia (modest), Vernalis (modest), SPOTRIAS NIH (significant); R. Gilberto Gonzalez: NIH Research Grant (significant), Berlex Consultant (modest); Raul G. Nogueira: Concentric Medical Inc, ev3 Neurovascular Inc, Coaxia, Inc. (all modest).
- Received January 20, 2009.
- Accepted April 17, 2009.
Schaefer PW, Barak ER, Kamalian S, Gharai LR, Schwamm L, Gonzalez RG, Lev MH. Quantitative assessment of core/penumbra mismatch in acute stroke: CT and MR perfusion imaging are strongly correlated when sufficient brain volume is imaged. Stroke. 2008; 39: 2986–2992.
Wintermark M, Albers GW, Alexandrov AV, Alger JR, Bammer R, Baron JC, Davis S, Demaerschalk BM, Derdeyn CP, Donnan GA, Eastwood JD, Fiebach JB, Fisher M, Furie KL, Goldmakher GV, Hacke W, Kidwell CS, Kloska SP, Kohrmann M, Koroshetz W, Lee TY, Lees KR, Lev MH, Liebeskind DS, Ostergaard L, Powers WJ, Provenzale J, Schellinger P, Silbergleit R, Sorensen AG, Wardlaw J, Wu O, Warach S. Acute stroke imaging research roadmap. Stroke. 2008; 39: 1621–1628.
Kudo K, Terae S, Katoh C, Oka M, Shiga T, Tamaki N, Miyasaka K. Quantitative cerebral blood flow measurement with dynamic perfusion ct using the vascular-pixel elimination method: comparison with h2(15)o positron emission tomography. AJNR. 2003; 24: 419–426.