Reversed Discrepancy between CT and Diffusion-Weighted MR Imaging in Acute Ischemic Stroke

Patients

Between May 1997 and June 2004, 215 consecutive patients who presented with acute MCA infarct and a National Institutes of Health Stroke Scale (NIHSS) score of more than 3 underwent CT within 6 hours of symptom onset. MR imaging was performed after CT on the same day in 210 patients. Of these 210, we retrospectively included 70 patients (40 men and 30 women; mean age, 70 years) who underwent MR imaging within 6 hours of symptom onset and who did not undergo thrombolytic therapy.

CT and MR Imaging

All CT scans were obtained using a helical CT scanner or multidector CT (Hi-Speed Advantage or LightSpeed Ultra; GE Medical Systems, Milwaukee, Wis). The scanning parameters of the unenhanced CT were 120 kV and 240 or 300 mAs with a 512 × 512 image matrix, 23- or 24-cm field of view, and a 5-mm section thickness. Maximal intensity projection CT angiography (CTA) was obtained from enhanced CT source data obtained according to a protocol of multiphasic perfusion CT.¹⁶

MR imaging was performed using a 1.5T unit (GE Medical Systems) with echo-planar imaging (EPI) sequences, including diffusion- and perfusion-weighted imaging (DWI/PWI). The typical stroke MR imaging protocol consisted of DWI, PWI, T2* gradient-echo, gadolinium-enhanced T1-weighted imaging, and 3D time-of-flight MR angiography (MRA). Imaging parameters of DWI were as follows: repetition time (TR), 6500–10,000 ms; echo time (TE), 71.7–96.8 ms; matrix, 128 × 128; 3 directions; field of view (FOV), 24 or 28 cm; section thickness, 5 mm; and intersection gap, 2 mm. DWI was obtained with b values of 0 and 1000 s/mm². PWI was performed with gradient-echo, echo-planar sequences with following parameters: TR, 2000 ms; TE, 60 ms; flip angle, 90°; matrix, 128 × 128, FOV, 24 cm; section thickness, 5 mm; and intersection gap, 2 mm. A series of images (8–10 sections, 40–50 images per section) was obtained before, during, and after administration of the contrast agent. Perfusion maps of relative cerebral blood volume (rCBV), time-to-peak (TTP), and relative cerebral blood flow (rCBF) were generated after eliminating the recirculation of contrast agent by γ-variate curve fitting. The rCBF map was obtained via the singular value decomposition deconvolution method.^18,19

Determination of Reversed Discrepancy, Extent of Ischemic Lesion and Arterial Occlusion

In a retrospective manner, 2 neuroradiologists (D.G.N. and E.Y.K.) independently reviewed CT and DWI and determined the presence of RD. Disagreements were decided by consensus. The interpreters were unaware of initial MRAs, follow-up images, and clinical information except for information of affected hemisphere. The lesion exhibiting RD was defined as an early ischemic lesion that showed parenchymal hypoattenuation or loss of gray and white matter distinction on CT but no obvious hyperintensity on DWI. RD was determined when the extent of the RD was equivalent to an Alberta Stroke Program Early CT Score (ASPECTS)²⁰ of at least 1 in the affected MCA territory. In all patients with RD, the extent of ischemic lesion was assessed using ASPECTS by parenchymal hypoattenuation on CT and by parenchymal hyperintensity on DWI. For determining ASPECTS, CT or DWI was reviewed independently of each image at a separate session. The 2 interpreters evaluated initial CTAs, initial MRAs, and follow-up (day 1) MRAs by consensus to determine arterial occlusion and presence of arterial recanalization, and recanalization rate of arterial occlusions at day 1 was compared between the patients with and without RD.

Quantitative Measurement of Reversed Discrepancy Lesions

In all patients with RD, initial and follow-up MR imaging, including DWI, PWI, and MRA, were successfully obtained 1 day after the initial MR imaging except for 1 patient (patient 4) in whom the initial PWI was not optimally obtained because of poor contrast enhancement. The PWI data from the patient 4 was not used in the analysis.

CT Hounsfield units (HU), ADC, DWI signal intensity, T2 signal intensity, TTP, rCBV, and rCBF values were measured by the final 3 regions of interest (ROIs) of RD lesion (only CT positive), DWI lesion 1 (both CT and DWI positive), and DWI lesion 2 (only DWI positive). All DWI and PWI images were spatially coregistered to the first volume of CT scans or EPI T2-weighted images (DWI b = 0) to superimpose the ROIs delineated on CT images using SPM2 (Wellcome Department of Cognitive Neuroscience). The 2 neuroradioligsts manually drew ROIs by consensus along the margins of the hypoattenuated lesions on CT and then drew ROIs along the margins of the hyperintense lesions on DWI in all patients with RD (Fig 1). Each ROI made on CT was superimposed on the same DWI images, which allowed us to make an ROI only for a lesion with RD (ROI_RD) and ROIs for the whole DWI lesions (ROI_DW), DWI lesion 1 (ROI_DW1), and DWI lesion 2 (ROI_DW2). The final 3 ROIs of RD lesion (ROI_RD), DWI lesion 1(ROI_DW1), and DWI lesion 2 (ROI_DW2) were transferred to the corresponding coregistered images of the initial CT, DWI, and PWI maps including TTP, rCBV, and rCBF. For comparison, mirror ROIs were drawn on the contralateral hemisphere. The size of ROI ranged from 1.2 to 30.4 cm². The ADC values were thresholded at 1200 × 10⁻⁶ mm²/s to minimize partial volume effect with CSF. The same quantitative measurements using ROIs were performed for follow-up MR imaging obtained at day 1 in patients with RD.

• Download figure
• Open in new tab
• Download powerpoint

Fig 1.

Patient 7. A, ROIs for parenchymal hypoattenuation and DWI hyperintensity were placed manually on unenhanced CT and DWI (TR, 10,000 ms; TE, 71.7 ms). The 2 ROIs for CT lesion and DWI hyperintensity provide an ROI for the lesion with reversed discrepancy showing CT parenchymal hypoattenuation and no obvious DWI hyperintensity.

B, Initial unenhanced CT shows subtle hypoattenuation of lentiform nucleus, caudate, and insula in the left MCA territory (white arrows). A hyperattenuated artery sign of the left inferior M2 branch (black arrow) is also seen. Maximum intensity projection CT angiography (CTA) shows occlusion of the left inferior M2 (arrow). DWI shows hyperintense lesions in the left temporal lobe and posterior insula, but DWI hyperintensity seems very subtle and not obvious in basal ganglia and is not seen in most areas of insula. ADC map shows cytotoxic edema in the left temporal lobe. Perfusion maps show an increased TTP delay and decreased rCBF only in the left temporal lobe, and no obvious perfusion abnormality is seen in the left basal ganglia. Initial MRA shows occlusion of the left inferior M2 (arrow).

C, Follow-up MR images at day 1 show delayed DWI hyperintensity and decreased ADC in the left basal ganglia (arrow). Perfusion maps and MRA demonstrate reperfusion in the left MCA territory and recanalization of the left M2 occlusion.

D, Follow-up CT and fluid-attenuated inversion recovery (FLAIR) images (TR, 10,000 ms; TE 133 ms; TI, 2200 ms) 7 days later show infarction in the RD lesion of basal ganglia and insula, and DWI shows increased extent of DWI hyperintense lesion in the left basal ganglia. FLAIR and T2-weighted MR images (TR, 3683 ms; TE, 104 ms) obtained 9 months after the initial CT show atrophy of the left basal ganglia and insula. T1 hyperintensity is seen in the left striatum at long-term follow-up. In this patient, there was a spontaneous rapid improvement of aphasia and motor weakness within 1 hour after symptom onset.

Clinical Data Analysis

A stroke neurologist assessed baseline, acute (days 1 and 7), and chronic (day 90) clinical data using the modified Rankin Score (mRS), and then compared for patients with and without RD. The incidence of spontaneous rapid clinical improvement (NIHSS score more than 3) within a few hours after stroke was assessed and compared for 2 subgroup patients. Time duration from onset to CT or MR imaging and the time interval between CT and MR imaging were compared between the 2 groups. The incidence of RD was compared between the patients who underwent MR imaging within 3 hours after symptom onset and those who underwent imaging at 3 to 6 hours. Hemorrhagic transformation was determined on T2* gradient-echo images at day 1.

Statistical Analysis

Statistical analysis was performed using commercial software (SPSS-PC, version 10.0; SPSS, Chicago, Ill). Comparisons of mean values for CT attenuation, ADC, DWI signal intensity, TTP, CBV, and CBF between lesions and contralateral hemispheres, and mean values for CT attenuation, ADC, ADC ratio, DWI signal intensity ratio, TTP delay, rCBV, and rCBF between lesions with RD and DWI lesions were compared with the use of a paired t test. When variables were non-normally distributed, comparisons were performed with the use of the Wilcoxon signed-rank test. Comparisons of clinical scores between the 2 subgroup patients with and without RD were performed with the use of the Student t test. The χ² test or Fisher exact test was used to compare recanalization rates of arterial occlusions at day 1 and the incidence of rapid clinical improvement between patients with and without RD. A P value of < 0.05 was considered statistically significant. Two observers independently determined ASPECT scores on both CT and DWI. Each observer determined the presence of RD when the case showed discrepancy in the areas of ASPECT score between CT and DWI. For the presence or absence of RD, interobserver agreement was evaluated with κ agreement index.

Demographic Data of Patients with Reversed Discrepancy

RD was found in 9 (12.9%) of 70 patients and located at basal ganglia in 8 (88.8%) of 9 patients with RD. The κ value for the presence of RD was 0.97. In all patients with RD, DWI hyperintense lesions were also present. Table 1 shows the demographic data of 9 patients (5 men and 4 women; mean age, 72 ± 12 years). The mean time to CT and MR imaging was 2.5 ± 1.3 and 4.1 ± 1.4 hours, respectively, after symptom onset, and mean time interval between initial CT and MR imaging was 1.6 ± 0.9 hours. A second follow-up CT or MR imaging was performed in 8 patients within 8 days (mean, 5.5 ± 2.4 days). No second follow-up imaging was performed in 1 patient (patient 4). In 4 patients, additional follow-up CT or MR images were obtained 3 to 11 months after the initial CT.

Extent of Ischemic Lesion and Arterial Occlusion

In 6 (66%) of the 9 patients with RD, ASPECTS of early ischemic lesion showing hypoattenuation on CT were lower than those on DWI, and mean ASPECTS of 9 patients with RD was significantly lower on CT than those assessed on DWI (4.6 ± 2.5 and 6.0 ± 2.1, respectively, P = .023). CTA was obtained in 8 of 9 patients with RD. CTA showed occlusion of M1 in only 3 patients and patent M1 segment in the other 4 of the 7 patients who had RD in basal ganglia. In 1 patient with RD in the frontal cortex, there was no arterial occlusion at CTA.

In 5 (56%) patients with RD, initial MRA vascular territory on the affected side was smaller than that of CT ischemic lesion (ie, mismatch). Follow-up MRA 1 day after the initial MRA demonstrated recanalization of arterial occlusions in 4 (50%) of 8 patients with RD who had arterial occlusions on the initial MRA. In 29 patients without RD who underwent both initial and follow-up MRA at day 1 and had arterial occlusions on the initial MRA, recanalization was observed in 7 patients on MRA at day 1. The recanalization rate of arterial occlusions at day 1 was higher in the patients with RD than in those without RD, but without significance (50% and 24%, respectively, P > .05).

Quantitative Measurement of Lesions Showing Reversed Discrepancy

Table 2 demonstrates mean values of CT HU, DWI, and PWI in lesions with RD on initial CT and MR imaging. The CT HU and TTP of RD lesion was significant different from contralateral control values (P < .001, respectively), but ADC and other MR imaging parameters of RD lesions were not different from control values. The mean CT HU of RD lesion was not significantly different from those of DWI lesion 1 (P > .05), but it was lower than that of DWI lesion 2 (P < .001). The mean ADC, ADC ratio, and DW signal intensity ratios of RD lesions were significantly different from those of DWI lesion 1 or 2 (P < .001). The mean TTP and rCBF values of RD lesions were less delayed and higher than those of DWI lesion 1 or 2 (P < 0.001, respectively). The mean rCBV value between RD lesion and DWI lesion 1 or 2 was not significantly different (P > .05).

Table 2 shows changes in mean ADC, DWI signal intensity, and T2 signal intensity values observed at the day-1 follow-up. For RD or DWI lesions, mean ADC values were significantly lower at follow-up than those at the initial DWI (P < .001), and DWI and T2 signal intensities were higher than those at the initial DWI (P < .001). T2 signal intensities of RD lesions were significantly lower than those of DWI lesion (P < 0.001) at the day-1 follow-up. The mean ADC value and DWI signal intensity of RD lesions were significantly different from those of DWI lesion (P = .037 and P < .001, respectively) at follow-up. The day 1 follow-up DWI demonstrated delayed hyperintensity of RD lesions in 8 (88.8%) of the 9 patients. In these 8 patients (except for 1 patient without a second follow-up image), all RD lesions progressed to overt infarctions at second follow-up. In 1 patient who had no delayed DWI hyperintensity at RD lesion at day 1, RD lesion progressed to infarction at the second follow-up. RD lesions converted to atrophy (n = 2) (Fig 1) or pan-necrosis (n = 2) (Fig 2) in 4 patients who underwent long-term follow-up CT or MR. The total area of CT hypoattenuation lesions and area of RD lesions was 500.25 and 128.47 cm², respectively. The area of RD lesion was present in 25.7% of total area of CT ischemic lesion in 9 patients with RD.

• Download figure
• Open in new tab
• Download powerpoint

Fig 2.

Patient 8. A, Initial unenhanced CT shows parenchymal hypoattenuation and loss of gray-white matter distinction in the left frontal lobe of ACA and MCA territory (arrows). Echo-planar spin-echo T2-weighted image (EPI T2, DWI of b value of 0, TR, 6500 ms; TE, 96.8 ms) shows no hyperintensity in the left frontal lobe. DWI (6500/96.8) and ADC map show hyperintensity and reduced ADC only in the medial frontal cortex and precentral gyrus that are only a part of hypoattenuated ischemic lesion depicted on CT. Perfusion maps show a TTP delay in the DWI lesion but no significant perfusion abnormality in the left frontal lobe lesion with reversed discrepancy on CT and DWI (arrowheads). MRA shows decreased signal intensity of the left distal MCA branches (arrow) and there was occlusion of the left A2 on MRA (not shown).

B, Follow-up MR images at day 1 show delayed DWI hyperintensity and ADC decrease in the left frontal lobe lesion with reversed discrepancy (arrows) and more severe cytotoxic edema in the medial frontal cortex and precentral gyrus of the left frontal lobe. Follow-up MRA shows recanalization of distal MCA branches but persistent perfusion abnormality in the ischemic lesion found on the initial DWI.

C, Follow-up CT image obtained 7 days after the initial CT demonstrates ischemic edema with heterogeneous hypoattenuation in the left frontal lobe. CT image obtained 11 months after the initial CT shows overt infarction with necrosis in the lesion with reversed discrepancy.

Clinical Data Analysis

No significant difference was observed in the baseline NIHSS scores (15.4 ± 4.5 and 13.6 ± 5.4), 24-hour NIHSS scores (13.7 ± 5.4 and 10.1 ± 6.2), or the chronic mRS (3.4 ± 1.6 and 2.8 ± 1.7) between those with and without RD on initial CT and DWI (P > .05). Spontaneous rapid clinical improvement was found in 3 (30%) patients with RD and 2 (3%) of patients without RD (P = .013). No significant difference was observed in terms of duration from symptom onset to CT and MR imaging (2.5 ± 1.3 versus 2.4 ± 1.2 and 4.1 ± 1.4 versus 3.8 ± 1.3, P > .05) or in terms of the time interval between CT and MR imaging (1.6 ± 0.9 and 1.4 ± 0.8, P > .05). No difference in the incidence of RD was found according to the time between onset and MR imaging, ie, 3 hours (13.0%, 3 of 23) and 3 to 6 hours after symptom onset (12.8%, 6 of 47) (P > .05). In 1 of 9 patients with RD, hemorrhagic transformation was observed in an RD lesion on T2* gradient-echo images at day-1 follow-up.