A retrospective analysis of negative diffusion-weighted image results in patients with acute cerebral infarction

We aimed to investigate the clinicoradiologic determinants of negative diffusion-weighted image (DWI) results in patients with acute cerebral infarction (ACI). The medical records were reviewed of ACI patients. Patients were divided to the DWI positive and negative group. Positive DWI was used as independent variable and patients' clinicoradiologic factors were used as co-variables for multivariate logistic regression analysis. 349 patients received initial cerebral MRI within 72 hours of admission. Lacunar infarction was most common (42.1%) followed by posterior circulation infarction (30.1%) and partial anterior circulation infarction (18.1%). The majority of the patients (72.2%) had an NIHSS score of less than 5 at admission. 316 patients (90.54%) were positive on initial DWI. Patients with smoking, initial SBP ≥ 140 or DBP ≥ 90 mmHg, initial fasting plasma glucose (FPG) ≥7.0 mmol/L, initial MRI from onset of disease >1 d and anterior circulation infarction were liable to show positive DWI. Furthermore, DWI negative patients had significantly lower NIHSS scores (IQR 0,1,2) than DWI positive patients (IQR 1,2,4) (P = 0.000) at two weeks post onset of acute cerebral infarction. In conclusion, multiple clinicoradiologic factors are associated with negative and positive DWI and further delineation of these factors is required in future prospective studies.

University and patient consent was not required due to the retrospective nature of this study. Patient information was anonymized and re-identified prior to analysis.
Patient evaluation. All patients were studied at admission as baseline evaluation. Patients' age, gender, time from symptom onset to first MRI, severity of neurological deficit on admission and stroke cause were retrieved. History of hypertension, diabetes, atrial fibrillation, smoking, and stroke was obtained. Initial blood pressure (BP) was measured on admission. Routine biochemical studies were done for fasting plasma glucose (FPG), glycated hemoglobin, initial hemoglobin, low density lipoprotein (LDL), and homocysteine. Severity of neurological deficit was evaluated at admission and two weeks post onset using the National Institutes of Health Stroke Scale (NIHSS), a 15-item scale that measures the level of neurologic impairment (scores range from 0 to 42, with higher scores indicating increasing severity). Examiners were trained stroke neurologists. Clinical classification was performed by trained stroke neurologists within one week after stroke onset using the Oxfordshire Community Stroke Project (OCSP) classification 11 and TOAST classification.
Imaging studies. MRI was done using Philips 3.0 T scanner. Cerebral MRI was done using T1WI, T2WI, DWI, and FLAIR. Scan was done with a slice thickness of 6 mm and at a slice interval of 1 mm with DWI b value 5 0, 1000 s/mm 2 . The scan parameters are shown in Table 1. Patients with negative DWI on initial MRI were reexamined one week after initial cerebral MRI, and the imaging data from the two examinations were compared to determine the presence of culprit lesions in combination with clinical manifestations. Patients were divided to the DWI positive and negative group according to initial DWI findings. The radiological data were interpreted by two equally experienced radiologists with more than 15 years of experiences in neuroimaging and there was no difference between their interpretations.
Therapeutic intervention. All patients with acute cerebral infarction received thrombolytic therapy within 3-4.5h and patients with posterior infarction received such therapy within 6 h of onset after exclusion of contraindications. Patients who were not indicated for thrombolysis were given antithrombotic and anticoagulation therapy.
Statistical analysis. Data were expressed as X+s or interquartile range (IQR) and analyzed using SPSS 17.0. Univariate analysis was done first and data were analyzed using Student's t test or Mann-Whitney U test. Chi square test or Fisher exact test was used for numerical data. Variables with P , 0.15 in the univariate analysis was entered into multivariate analysis. In addition, OCSP and TOAST classification, and baseline NIHSS scores were entered into multivariate logistic regression analysis using positive DWI as independent variable and the clinicoradiologic factors were used as co-variables. P , 0.05 was considered statistically significant.

Results
Patient demographic and baseline characteristics. Four hundred and eighty-four patients were hospitalized because of acute cerebral infarction during the review period and 349 of them received initial cerebral MRI within 72 hours of admission. The study flowchart is shown in Fig. 1 and patient demographic and baseline characteristics are listed in Table 2. The time to initial cerebral MRI from onset of disease was 24 hours or less in 18.3% of the patients. Lacunar infarction was most common in these patients (42.1%) followed by posterior circulation infarction (30.1%) and partial anterior circulation infarction (18.1%). The majority of the patients (72.2%) had an NIHSS score less than 5 at admission.
Characteristics of DWI positive and negative patients with acute cerebral infarction. Three hundred and sixteen (90.54%) patients were positive and 33 (9.46%) were negative on initial DWI. Their demographic and baseline data are shown in Table 2. A greater proportion of DWI positive patients were male (63.6% vs. DWI negative, 45.5%; P . 0.05). A greater proportion of patients had higher initial BP (77.8% vs. DWI negative, 66.7%; P . 0.05) and initial FPG (26.9% vs. DWI negative 9.1%; P , 0.05) in patients with positive DWI. A higher proportion of patients who were DWI negative underwent initial cerebral MRI within 24 hours of disease onset (30.3% vs. DWI positive, 17.1%; P . 0.05). Though a smaller percentage of DWI positive patients had an NIHSS score less than 5 at admission (71.2%), there was no statistical difference compared to DWI negative patients (81.8%; P . 0.05). DWI negative and positive patients were comparable in other demographic and baseline characteristics ( Table 2). According to the OCSP classification, approximately half (51.5%) of DWI negative patients had lacunar infarction followed by posterior circulation infarction (42.4%) (Table 3). Similarly, DWI positive patients, 41.1% had lacunar infarction followed by (posterior circulation infarction (28.8%). According to the TOAST classification, 60.6% of DWI negative patients (vs. 41.8% in DWI positive patients) had small artery occlusion followed by large artery arteriosclerosis (33.3% vs. 46.7% in DWI positive patients).
Moreover, five (15.2%) patients who were DWI negative initially showed positive findings upon reexamination one week later; three   patients showed lesions on both DWI and FLAIR while two patients showed lesions on FLAIR only ( Fig. 2 and 3).

Risk factors for positive DWI findings upon initial cerebral MRI.
Our univariate analysis showed that smoking (P 5 0.026), baseline BP (P 5 0.148), baseline FPG (P 5 0.025), LDL (P 5 0.037) and time to initial cerebral MRI (P 5 0.062) were significantly different between DWI positive and negative patients ( Table 2). These factors were further used as covariates for multivariate logistic regression analysis using positive DWI as an independent variable. We found that history of smoking was associated with a 2.8-fold increase in the risk for positive DWI (P 5 0.0196; OR: 2.772; 95%CI, 1.177-6.525) (   [12][13][14] . The brain maintains apparently intact morphology when perfusion is reduced to 30% of normal though neurons may partially lose certain function. If perfusion is returned to normal, no abnormality will be detected by cerebral MRI. However, if hypoperfusion worsens and cerebral blood flow is further reduced to 15-20% of normal, the function of membrane Na-K-ATPase is impaired, leading to disequilibrium of ions across the membrane and the development of cytotoxic edema 15 . High intensity signal appears in DWI early thanT1W, T2W and FLAIR. Negative DWI on initial cerebral MRI in acute cerebral infarction patients has been reported, but most of them are anecdotal case reports [4][5][6][7] . In our study, we used 3.0 T, which avoids the impact of magnetic field strength on DWI. Rosso et al. 8 reviewed the imaging data of patients within 6 hours of anterior circulation infarction by 3.0 T and 1.5 T MR and found that 3.0 T DWI had a higher false positive rate. However, unlike our study, they only included patients with anterior circulation infarct and no patients with posterior infarction or lacunar infarction. Here, our retrospective analysis of 349 patients showed that close to one in ten acute cerebral infarction patients (9.5%) showed negative DWI on initial cerebral MRI. This is higher than the rate (5.8%) reported by Oppenheim et al. 4 , but lower than that (25.6%) reported by Sylaja et al. 16 . Time to initial cerebral MR was .3 hours from onset for our patients; patients who were within the time window for thromoblysis therapy were DWI positive or DWI negative. Those who were DWI negative on initial cerebral MR still received thromoblysis after contraindications were ruled out. Because these patients' symptoms persisted less than 24 hours, it remains uncertain whether they had transient ischemic attack (TIA). It has been reported that positive DWI in TIA patients correlated with symptoms that persist $60 min and the more prolonged the symptoms were, the greater the likelihood of permanent pathological injury was 17,18 . This is also the reason for modified definition of TIA. For those who had initial cerebral MR within 24 hours of disease onset and were still DWI negative on repeat cerebral MR, we believe that these patients had cerebral infarction as their neurologic deficits persisted more than one week. For these patients, perfusion weighted imaging may help define cerebral infarction or TIA. In addition, Qiao et al. reported that spin labeling may help identify perfusion defects in TIA patients 19 .    Clinical correlates of initial positive DWI in patients with actue cerebral infarction and the underlying mechanisms. We found that, compared to DWI negative patients, a greater proportion of DWI positive patients were male, had higher initial BP and initial FPG. In addition, a markedly greater proportion of DWI positive patients had a lower rate of posterior circulation infarction, but a higher rate of anterior circulation infarction. DWI negative and positive patients were comparable in other demographic and baseline characteristics. Our multivariate logistic regression analysis showed that history of smoking, initial SBP $ 140 or DBP $ 90 mmHg, initial FPG $ 7.0 mmol/L, initial MRI from onset of disease .24 hours and anterior circulation infarction were associated with markedly increased risk for positive DWI. We revealed it is interesting that smoking markedly increased the risk of positive DWI on initial cerebral MRI (P , 0.05). Smoking is known to be closely associated with ischemic stroke. Abbruscato et al. reported that the activity of Na-K-2Cl cotransporter is increased to maintain ion equilibrium across the cells in the blood brain barrier when the activity of Na-K-ATPase is decreased as occurs during cerebral ischemia 20 . Paulsonet et al. showed that nicotine markedly decreased the activity of Na-K-2Cl cotransporter, leading to cellular toxic edema 21 , which may explain why cerebral infarction patients with a history of smoking are at increased risk for positive DWI.
It has been further shown that increased FPG at admission is a negative predictor for the outcome of acute ischemic stroke patients 22,23 . Our multivariate logistic analysis showed no statistical difference in our DWI negative and positive patients in history of diabetes, and glycated hemoglobin, but revealed that initial FPG $ 7.0 mmol/L was a risk factor for positive DWI in acute cerebral infarction patients. Though a similar proportion of acute cerebral infarction patients had a history of hypertension in the DWI negative and positive group, we found that higher initial BP ($140/ 90 mmHg) was a risk factor for positive cerebral DWI. Clinical and animal studies showed that higher initial FPG induces damages against the vascular endothelium and increases permeability of blood brain barrier, consequently enlarging the size of the infarct and aggravating edema [24][25][26] . Hou et al found that high BP increased the amount of G-strophanthin analogues in the brain that inhibit Na 1 K 1 ATPase 27 . We speculate that these could be one of the causes that higher initial FPG and BP are risk factors for positive DWI in acute cerebral infarction patients.

Effect of time to initial cerebral MRI and infarct location on DWI.
We found that site of infarction and time to initial cerebral MRI correlated with negative DWI in acute cerebral infarction patients. Our multivariate analysis showed that time to initial cerebral MRI is an important risk factor for positive DWI with time to initial MRI within 24 hours of disease onset more likely to yield negative DWI lesions. This is consistent with the findings by Oppenheim et al. who found that negative DWI finding correlated with time to initial MRI examination in patients with acute cerebral infarction 4 . The earlier the time to initial cerebral MRI, the more likely it is to obtain negative DWI findings. Five (15.2%) of our patients who were initially DWI negative showed positive findings upon reexamination one week later.
Current literature indicates negative DWI on initial cerebral MRI also correlates with site of cerebral infarction; Negative DWI lesions are seldom reported in patients with anterior circulation infarction 6 . In the current study, 2 patients who were initially DWI negative had anterior circulation infarction and they received initial cerebral MRI within 4.5 hours of disease onset. Our multivariate logistic regression analysis revealed that anterior circulation infarction patients are more likely to show positive DWI than posterior circulation infarction patients. Posterior circulation infarction patients may exhibit imaging abnormalities in DWI later in the acute phase compared with other sites of infarction. Periodic reexamination may help yield positive DWI findings. Posterior circulation infarction exhibits imaging abnormalities in DWI later than infarction at other sites [28][29][30][31] .  Infarct is small in the posterior circulation and may not become detectable even on repeat cerebral MRI. It has been reported that there was a gap in time between the development of neuronal dysfunction due to focal ischemia and water diffusion difficulty, which may cause DWI delay. Another contributor to delayed DWI positivity is inadequate signal to noise ratio in the early stage of the disease or MR artifacts interfering with DWI. Some investigators speculated about the presence of a second ischemic episode [4][5][6][7][28][29][30][31] . This hypothesis, however, is not supported by the findings of the current study which showed improvement of neurological function of DWI negative patients between two MRI examinations. Sylaja et al. reexamined DWI negative patients 30 days after initial MRI of acute cerebral infarction patients and found that a portion of these patients were still DWI negative even though hyperintense signal was observed in FLAIR 16 . Some of our DWI negative patients classified with lacunar infarction and posterior circulation infarction who did not receive thrombolytic therapy showed negative MRI findings even though their symptoms persisted more than 24 hours. Therefore, for DWI negative patients, prompt and periodic reexamination by MRI may identify the culprit lesion in acute cerebral infarction patients.
Future directions and limitations. Mismatch between DWI and perfusion weighted imaging (PWI) is widely believed to be an indication of salvageable tissue after an infarction 35 . The DWI abnormality is typically smaller than the PWI abnormality, and the region where both are abnormal is regarded as infarct core, and unsalvageable. The area where PWI is abnormal but DWI is normal, the penumbra, is typically salvageable. In DWI negative patients, a PWI lesion might be visible to highlight the area of infarction. Furthermore, this could indicate that this area is not permanently damaged and that reperfusion would strongly benefit this area. This is a commonly used measure in stroke imaging 33 . In this retrospective study, PWI was not performed when DWI was carried out and therefore we could not determine whether there was a mismatch between PWI and DWI. This is a limitation of the current study. Cho et al. reported that 3 patients with acute cerebral anterior circulation infarction showed negative DWI on cerebral MR while PWI indicated the presence of perfusion defect 34 . Fustier et al. reviewed DWI negative patients on initial cerebral MRI who received thromoblysis therapy and found that these patients had better prognosis than DWI positive patients 35 . Diffusion-perfusion mismatch helps determine early whether patients can benefit from thromoblysis therapy 36 . In DWI negative patients, the following scenarios may occur if PWI is undertaken: 1) there is a total mismatch between the two (PWI1 and DWI2). Ischemia is demonstrable on one hand while there are no positive DWI findings on the other hand after prompt reperfusion. 2) PWI and DWI are both negative, which is frequently seen in migraine and lasts for 1-48 hours 32 . Straka et al. reported a real time evaluation system which can detect the diffusion-perfusion mismatch in 5 to 7 minutes 33 , which may facilitate our exploration of the causes and mechanisms of negative DWI.

Conclusion
In conclusion, we confirmed that negative DWI is associated with time to initial cerebral MRI after disease onset and site of infarction. The earlier the time to initial cerebral MRI is, the more likely a negative DWI is. Furthermore, posterior circulation infarction patients are more likely to have negative DWI than anterior circulation infarction patients. Patients with history of smoking, higher initial BP and FPG are more likely to have positive DWI lesions. However, the current study is limited by the retrospective nature of the analysis and the conclusion may not be applicable to primary healthcare settings as it was carried out at a single tertiary care institution. In addition, even though patients were included who received initial cerebral MRI within 72 hours of admission, great variation may exist in time to initial cerebral MRI due to variation in time to seek medical care after disease onset. Furthermore, the majority of our patients had NIHSS , 5 and had mild stroke and the number of DWI negative patients was relatively low. Another limitation of the current study is the lack of a gold standard for the ultimate diagnosis of acute cerebral infarction even though DWI negative patients in the study met the clinical definition for cerebral infarction. This will be addressed in our future studies by using PWI or adjusting b value in DWI. Therefore, the findings of the current study need to be confirmed by prospective studies with more clearly defined time to initial cerebral MRI from disease onset.