Temporal and Spatial Profile of Brain Diffusion-Weighted MRI After Cardiac Arrest
Background and Purpose— Diffusion-weighted magnetic resonance imaging of the brain is a promising technique to help predict functional outcome in comatose survivors of cardiac arrest. We aimed to evaluate prospectively the temporal-spatial profile of brain apparent diffusion coefficient changes in comatose survivors during the first 8 days after cardiac arrest.
Methods— Apparent diffusion coefficient values were measured by 2 independent and blinded investigators in predefined brain regions in 18 good- and 15 poor-outcome patients with 38 brain magnetic resonance imaging scans and were compared with those of 14 normal controls. The same brain regions were also assessed qualitatively by 2 other independent and blinded investigators.
Results— In poor-outcome patients, cortical structures, in particular the occipital and temporal lobes, and the putamen exhibited the most profound apparent diffusion coefficient reductions, which were noted as early as 1.5 days and reached a nadir between 3 and 5 days after the arrest. Conversely, when compared with normal controls, good-outcome patients exhibited increased diffusivity, in particular in the hippocampus, temporal and occipital lobes, and corona radiata. By qualitative magnetic resonance imaging readings, 1 or more cortical gray matter structures were judged to be moderately to severely abnormal in all poor-outcome patients except for the 3 patients imaged within 24 hours after the arrest.
Conclusions— Brain diffusion-weighted imaging changes in comatose, postcardiac arrest survivors in the first week after the arrest are region and time dependent and differ between good- and poor-outcome patients. With increasing use of magnetic resonance imaging in this context, it is important to be aware of these relations.
Cardiopulmonary arrest is associated with high morbidity and mortality1–3 that are often caused by hypoxic-ischemic brain injury. Therefore, the ability to accurately assess the presence and severity of brain injury early after the arrest is a critical issue for healthcare providers and patients’ families.
Brain diffusion-weighted (DWI) magnetic resonance imaging (MRI) is exquisitely sensitive in the detection of early ischemic brain injury.4,5 In the presence of acute ischemic infarction, the apparent diffusion coefficient (ADC) decreases by 25% to 40% in the affected territories within minutes to hours after symptom onset, reaching its nadir at ≈24 hours and remaining decreased for 7 to 14 days.6
Brain regions with reduced diffusion also occur in patients with severe global hypoxic-ischemic brain injury after cardiac arrest and are correlated with outcome in preliminary reports.7–15 In a previous study, >10% of brain volume with an ADC value <650×10−6 mm2/s to 700 ×10−6 mm2/s between 2 and 4.5 days after the arrest identified poor-outcome patients with 100% specificity and 81% sensitivity.14 The timeline and the spatial profile of these DWI changes, however, have not been well defined. In this study, we aimed to define the temporal and spatial profile of brain ADC changes caused by global hypoxic-ischemic brain injury in specific and predefined brain regions during the first 8 days after cardiac arrest.
Consecutive comatose, postcardiac arrest patients were prospectively enrolled. Details of the study methodology have been published previously.14 In brief, adult patients who remained comatose after successful resuscitation for cardiac arrest were eligible. Patients were excluded when they had a preexisting severe coexisting systemic disease, “do not resuscitate” status, or a baseline modified Rankin Scale score of ≥3. Patients underwent standardized neurologic examinations at 1 hour and at 1, 2, and 3 days after the arrest. Somatosensory evoked potentials were performed at 3 days after the arrest if the patients were still comatose. The study was approved by the institutional review board, and written consent was obtained from legally authorized representatives for each participant.
Patients were included in this study if they underwent a technically adequate brain DWI study within 8 days of the arrest and if they either survived for 6 months or died while meeting any of the following predefined clinical criteria shown to be highly specific for poor outcome: absent pupillary reflexes or motor response at 3 days, bilateral absent cortical responses by somatosensory evoked potentials after 2 days, and/or vegetative state after 1 month. Patients who died after withdrawal of life support or because of a cardiopulmonary complication without meeting any of these specific clinical criteria were not included in this report because their outcomes were considered indeterminate. In patients treated with induced hypothermia, MRIs were obtained after passive rewarming to normal body temperature. Good versus poor outcomes were defined as a 6-month Glasgow outcome scale (GOS) score of ≥3 versus 1 or 2.
MRI Protocol and Image Processing
The MRI protocol for this study has been previously reported.14 In brief, DWI images were acquired on a 1.5-T GE Signa Horizon scanner (GE Medical Systems) with spin-echo echo-planar imaging, 128×128 acquisition matrix, 256×256 reconstructed matrix, field of view 240×240 mm, slice thickness/gap of 5/1.5 or 5/2.5 mm, b=0 and 1000 seconds/mm2, and diffusion encoding along the principal axes averaged to provide isotropically-weighted DWIs. Images were processed and analyzed with software developed in house (Diffusion/Perfusion Analysis, UCLA Stroke Center, Stanford Stroke Center). ADC maps were created. Two independent and blinded investigators manually outlined predefined regions of interest (ROIs) bilaterally (except for midline structures) in the frontal, parietal, occipital, and temporal lobes; hippocampus; corona radiata; internal capsule; caudate; putamen; thalamus; cerebellum; and pons (Figure 1). The program calculated the mean and standard deviation of the ADC value in the outlined ROIs.
All brain MRIs were also read systematically and qualitatively by a neuroradiologist and a neurologist with subspecialty certification in neurocritical care and vascular neurology. Evaluators were independent and blinded to patient identity and outcome. The fluid-attenuated inversion recovery (FLAIR) and DWI sequences were scored systematically on the basis of the severity of signal abnormality attributed to the cardiac arrest in 15 predefined regions corresponding to the regions included in the quantitative analyses: cortical gray and subcortical white matter in the frontal, parietal, temporal, and occipital lobes; hippocampus; caudate nucleus; putamen; thalamus, cerebellum (cortex and white matter); and pons. All brain regions were scored according to the severity of signal abnormality on a 5-point scale with 0=no abnormality, 1=possibly abnormal, 2=mildly abnormal, 3=moderately abnormal, and 4= severely abnormal. The average scores of the 2 raters were used.
Brain MRIs of 14 patients without evidence of acute infarction, without other acute parenchymal findings on their brain MRIs, and without a diagnosis of cerebrovascular disease were used as controls. The reasons for obtaining MRI in these 14 patients were nonspecific and transient neurologic symptoms including the following: headache, numbness, generalized weakness, altered mental status, and transient visual changes. The discharge diagnoses of these 14 patients were migraine with aura or migraine equivalents (n=4), presyncope (n=2), sensory neuropathy (n=1), catatonia (n=1), polymyalgia rheumatica (n=1), perimesencephalic (nonaneurysmal) subarachnoid hemorrhage (n=1), systemic infection with altered mental status (n=1), and uncertain (n=3).
MRIs were assigned to 48-hour time intervals after cardiac arrest. Each interval overlapped by 12 hours with the adjacent ones. Scans were assigned to 2 intervals when they were done during the overlapping time window. Mean ADC values of the ROIs in the various brain structures of each patient within each time interval in the good- versus poor-outcome groups were compared with the control group by the Mann–Whitney U test. All statistical tests were 2 tailed, and significance was defined at α<0.05. To control for multiple comparisons, we adjusted the false discovery rate q*=0.05 by using the step-up controlling procedure of Benjamini and Hochberg.16 Statistical analyses were performed with SPSS 17.0 (SPSS Inc).
Twenty brain MRIs of 18 good-outcome patients and 18 MRIs of 15 poor-outcome patients were analyzed. The FLAIR sequence of 1 patient could not be interpreted qualitatively because of motion artifact. MRIs were obtained at a median (interquartile range) of 80 (55–117) hours after cardiac arrest. Baseline characteristics of the good- and poor-outcome patients are presented in Table 1. Eleven patients in the good- and 10 in the poor-outcome group underwent induced hypothermia. The 15 poor-outcome patients met the following clinical criteria for unfavorable outcome: absent pupillary reflexes at 72 hours (n=5), absent motor response at 72 hours (n=12), absent somatosensory evoked potentials after 48 hours (n=10), and inability to regain consciousness at 1 month (n=1). In the good-outcome group, the 6-month GOS was 3 in 6, 4 in 5, and 5 in 7 patients. The control group consisted of 8 women and 6 men with a mean age of 45±12 years. Baseline characteristics did not differ among the 3 groups.
Qualitative MRI Readings
Qualitative MRI readings are presented in Table 2 and Figure 2. In the good-outcome group, all but 4 patients were judged to have normal cortical structures. Four patients were thought to have mild to moderate cortical gray matter abnormalities on the DWI sequence, the FLAIR sequence, or both. Of these, 1 patient had a GOS=3, 2 a GOS=4, and 1 a GOS=5 at 6 months.
In the poor-outcome group, the cortical gray matter structures were the most severely affected of all ROIs. One or more of these structures were thought to be moderately or severely abnormal on the DWI or FLAIR sequences in all but 3 patients. Two patients who were scanned very early (at 2 and 7 hours after the arrest) were judged to have a normal cortex, and 1 patient, who was scanned at 14 hours, had only mild DWI signal abnormalities in his cortical gray matter structures. Generally, the DWI sequence was read as being more abnormal than the FLAIR sequence.
Moderate to severe DWI and FLAIR signal abnormalities were seen in 4 patients in the good-outcome group and in more than half of the patients in the poor-outcome group.
Deep Gray Nuclei
In the good-outcome group, scans of half of the patients were read as having normal deep gray nuclei. However, the other half were thought to have mild to moderate abnormalities on DWI and FLAIR in these structures. In the poor-outcome group, the caudate, putamen, and thalamus were mildly to severely affected on DWI and FLAIR sequences in all patients, except for the 2 patients who were scanned at 2 and 7 hours. Most patients had moderate to severe abnormalities in these structures.
Cerebellum and Pons
Brainstem structures were thought to be normal in all good- and poor-outcome patients with the exception of 3 poor-outcome patients who were judged to have moderate abnormalities on both DWI and FLAIR sequences. Furthermore, more than half of the poor-outcome patients had moderate to severe cerebellar abnormalities on DWI and FLAIR that affected the cerebellar cortex but not the cerebellar white matter. Significant signal abnormalities in the brain stem and cerebellum did not occur in the good-outcome group except for 1 patient who had severe abnormalities of the cerebellar cortex on both DWI and FLAIR sequences.
Quantitative MRI Measurements
The reproducibility of the quantitative ROI measurements was excellent (intraclass correlation coefficient of 0.923 and 0.935 for interrater and intrarater reliability, respectively). The median of the mean ADC value for each ROI in each time interval is presented in Table 3⇓ for good- and poor-outcome patients and was compared between the 2 groups as well as with normal controls.
In the good-outcome group, the ADC values of the temporal and occipital lobes were increased compared with those in controls starting at 1.5 days, a trend that persisted throughout the first week after the arrest. In contrast, poor-outcome patients exhibited early and marked reduced diffusion in the frontal, parietal, temporal, and occipital lobes starting at 1.5 days, which reached a nadir between 3 and 5 days (Figure 3A). The ADC changes were most profound in the occipital cortex.
In the good-outcome patients, the hippocampus demonstrated increased diffusivity starting in the early time intervals. In poor-outcome patients, there was a trend toward a decline in ADC values in the 3- to 6.5-day time interval, but the ADC values did not significantly differ from those of controls or of good-outcome patients in the individual time windows (Figure 3B and Table 3⇑).
Deep Gray Nuclei
The deep gray matter structures (thalamus, caudate, and putamen) in the good-outcome patients demonstrated an early trend toward increased diffusivity that reached significance only in the putamen between 3 and 5 days. Again, the opposite trend was noted in the poor-outcome group, with reduced diffusion in the first 5 days, reaching significance in the putamen in the 1.5- to 3.5-day interval (Figure 3C). Overall, reduced diffusivity in poor-outcome patients was more severe in cortical structures than in the deep gray nuclei.
Corona Radiata and Internal Capsule
In good-outcome patients, the corona radiata but not the internal capsule showed mild increased diffusion in all time intervals starting at 1.5 days, with a peak between 3 and 5 days. In contrast, in the poor-outcome group, marked reduced diffusion was noted in the corona radiata and internal capsule only between 6 and 8 days (Figure 3D).
Cerebellum and Pons
In the good-outcome group, the cerebellum exhibited increased diffusivity most markedly in the 3- to 5-day interval. Diffusivity in the cerebellum of the poor-outcome patients tended to be decreased compared with controls starting at 1.5 days and was most pronounced between 6 and 8 days. The pons demonstrated a trend toward mild increased diffusion at most time points in both groups but more so in the good- than in the poor-outcome patients. Finally, similar to the qualitative findings, poor-outcome patients had significantly lower ADC values in each structure than did good-outcome patients when all time windows were combined (ie, 0 to 192 hours), except for the internal capsule, pons, and caudate; a trend towards lower ADC values was observed in the thalamus.
DWI detects early cytotoxic edema by measuring the random motion of water protons, a process that is reduced by failure of the energy-requiring active water transport mechanism. DWI studies in animal models have demonstrated a decline in brain ADC values during cardiac arrest, which then reverse after successful resuscitation.17–19 Despite successful reperfusion, however, secondary energy failure and ADC decrease follow after several hours.20
It has been shown in preliminary studies that quantitative MRI brain changes are correlated with functional outcome in comatose postcardiac arrest survivors.14,21 The percentage of brain tissue below a threshold of 650×10−6 mm2/s to 700×10−6 mm2/s was found to be correlated with functional outcome at 3 months after the arrest.14 Based on whole-brain quantitative DWI analyses, the ideal time window for prognostication appears to be between 49 and 108 hours after the arrest, when the ADC reductions are most apparent. None of the patients with >10% of brain tissue with an ADC value <650×10−6 mm2/s to 700×10−6 mm2/s during this time window regained consciousness.
In this prospective study, we analyzed in detail which brain structures at which time point were the most severely affected by ADC reductions by comparing good- and poor-outcome patients with normal controls and with each other. Our results show that ADC changes caused by global ischemic brain injury in humans are very delayed compared with ADC changes caused by focal cerebral ischemia. In severe global ischemic brain injury associated with poor outcome, reduced diffusion may not be apparent in the initial hours after the arrest.
Because physicians are increasingly using brain MRI with DWI for prognostic purposes in postcardiac arrest patients, it is important to be aware that ADC changes in these patients are both time and region dependent during the first week. We found that both the qualitative and quantitative MRI changes in poor-outcome patients were most severe in the cortical regions and most apparent between 3 and 5 days after the arrest. Although ADC changes occur globally, they most profoundly affect the cortical gray matter structures in the poor-outcome patients in the first week after the arrest. None of the patients with moderate to severe cortical abnormalities on their MRI awoke from their coma. Importantly, the 3 poor-outcome patients who did not display substantial cortical abnormalities on their DWI MRI were imaged at 2, 7, and 14 hours after their arrests, suggesting that these changes are not apparent early after the arrest. In contrast to the gray matter structures, reduced diffusion in the white matter structures was not observed until the end of the first week. This DWI pattern likely reflects differences in tissue response to ischemic injury between various brain structures.22
Most patients who regained consciousness had scans that were qualitatively read as having normal cortical structures, with the exception of 4 patients who had mild to moderate abnormalities on FLAIR and/or DWI sequence in 1 or more cortical gray matter structures. Furthermore, half of the good-outcome patients had qualitative changes in the deep gray nuclei. Thus, mild to moderate cortical abnormalities and abnormalities in the deep gray nuclei did not exclude the possibility of regaining consciousness. Because qualitative interpretation of symmetrical and potentially subtle MRI changes is likely to vary between observers, we suspect that quantitative ADC changes may be more specific for prognostication.
Prior data on the temporal and spatial profile of brain ADC changes in the first week after cardiac arrest are scarce.21,23 A recently published retrospective study confirms our observation that regional brain ADC values differ between good- and poor-outcome cardiac arrest patients and are time dependent.21 Comparisons with normal controls were not performed. To our surprise, when compared with normal controls, we observed increased or facilitated (rather than reduced or restricted) diffusion in good-outcome patients who awoke from their coma. This change was only apparent after our quantitative ADC analyses and most pronounced in the cortical structures, hippocampus, putamen, and corona radiata. Increased diffusion has, to our knowledge, not been reported previously in humans in this context. We hypothesize that this increased diffusivity may represent mild global vasogenic edema caused by transient increased blood-brain barrier permeability.
Although diffusivity in the hippocampus of good-outcome patients was consistently increased, no difference was observed in the poor-outcome patients compared with controls. One possible explanation is that severe hippocampal hypoxic-ischemic injury is associated with both vasogenic and cytotoxic edema, which counterbalance each other’s effects on the hippocampal ADC values.
The differences in DWI data of individual structures between good- and poor-outcome patients were very similar in the quantitative and qualitative analyses when all time windows were combined. The only exception was the caudate nucleus, which was rated distinctly different in the 2 groups with qualitative DWI readings but not with the quantitative measurements. The most likely explanation for this observation is that there is increased T2 signal effect in the DWI maps easily visualized qualitatively but not measured quantitatively as reduced diffusion.
An important limitation of this study is a limited number of MRI observations in each time window, in particular, in the very early time window of 0 to 48 hours. The main reason why patients were not scanned during this time was because they were undergoing therapeutic hypothermia, and we wanted to avoid any possible effect of hypothermia on the quantitative DWI MRI results. In addition, patients who were taken off life support without meeting specific clinical criteria for poor outcome were excluded from this study. We decided not to include these patients because their outcomes were considered uncertain. Lastly, the results of this study only apply to patients who underwent brain MRI within 8 days after cardiac arrest. We are aware that some patients are too unstable from a critical care standpoint to undergo MRI during this early time period. MRI changes caused by severe hypoxic-ischemic brain injury are different after the first week.7
In summary, we found that moderate to severe reduced diffusion in cortical regions is strongly associated with the inability to regain consciousness and that this finding is most apparent on MRI between 3 and 5 days after the arrest. In contrast, patients who are able to wake up from their coma exhibit increased diffusion involving the temporal and occipital lobes, corona radiata, and hippocampus and qualitative changes in the deep gray nuclei alone are common in these patients. Because brain MRI is increasingly used for prognostic purposes in critically ill neurologic patients, physicians need to be aware that brain diffusion changes are both time and region dependent in the context of diffuse hypoxic-ischemic brain injury during the first week.
This study was funded by the National Institutes of Health 1R01HL089116 (to Dr Wijman), R01NS034866 (to Dr Moseley), 2R01EB002711 (to Dr Bammer), the American Heart Association National Scientist Development Award 0430275N (to Dr Wijman), and the Katherine McCormick Fund for Women (to Dr Hsia). The authors would like to thank Stephanie Kemp for administrative support of this study and Anna Finley Caulfield, Marion Buckwalter, Chitra Venkatasubramanian, Maarten Lansberg, and Neil Schwartz for their assistance with patient enrollment.
Sources of Funding
This study was funded by the American Heart Association (AHA) and the National Institutes of Health (NIH). Dr Bammer was funded by NIH grant No. 2R01EB002711. Dr Hsia received a stipend from the Katherine McCormick Fund for Women. Dr Moseley was funded by NIH grant No. R01NS034866. Dr Wijman is funded by NIH grants No. R01HL089116 and 2R01NS034866 and received the AHA National Scientist Development award No. 0430275N for this research.
Dr Leproust is employed by Agilent Technologies, Inc. The remaining authors report no conflicts of interest.
- Received February 21, 2010.
- Revision received April 4, 2010.
- Accepted April 27, 2010.
Moseley ME, Kucharczyk J, Mintorovitch J, Cohen Y, Kurhanewicz J, Derugin N, Asgari H, Norman D. Diffusion-weighted MR imaging of acute stroke: correlation with T2-weighted and magnetic susceptibility-enhanced MR imaging in cats. AJNR Am J Neuroradiol. 1990; 11: 423–429.
Arbelaez A, Castillo M, Mukherji SK. Diffusion-weighted MR imaging of global cerebral anoxia. AJNR Am J Neuroradiol. 1999; 20: 999–1007.
Wijdicks EFM, Campeau NG, Miller GM. MR imaging in comatose survivors of cardiac resuscitation. AJNR Am J Neuroradiol. 2001; 22: 1561–1565.
McKinney AM, Teksam M, Felice R, Casey SO, Cranford R, Truwit CL, Kieffer S. Diffusion-weighted imaging in the setting of diffuse cortical laminar necrosis and hypoxic-ischemic encephalopathy. AJNR Am J Neuroradiol. 2004; 25: 1659–1665.
de Crespigny AJ, Rother J, Beaulieu C, Moseley ME, Hoehn M. Rapid monitoring of diffusion, DC potential, and blood oxygenation changes during global ischemia: effects of hypoglycemia, hyperglycemia, and TTX. Stroke. 1999; 30: 2212–2222.
Krep H, Bottiger BW, Bock C, Kerskens CM, Radermacher B, Fischer M, Hoehn M, Hossmann KA. Time course of circulatory and metabolic recovery of cat brain after cardiac arrest assessed by perfusion- and diffusion-weighted imaging and MR-spectroscopy. Resuscitation. 2003; 58: 337–348.