Diffusion-weighted magnetic resonance imaging for predicting the clinical outcome of comatose survivors after cardiac arrest: a cohort study

This study was reviewed and approved by the local ethics committee of our university hospital. Between January 2004 and December 2007, we prospectively studied 39 patients at St. Mary's Hospital (a tertiary-care university hospital in Seoul, Korea) who survived an out-of-hospital cardiac arrest. We included the adult patients (≥ 18 years) who were successfully resuscitated from the cardiac arrest, survived for at least 24 h, and remained comatose for at least 6 h after return of spontaneous circulation (ROSC) to avoid transient unconsciousness. The exclusion criteria included cardiac arrest resulting from intracranial haemorrhage, drug intoxication, trauma or a terminal illness, a previous history of neurological disease or brain trauma, a lack of informed consent, and being unavailable for follow-up. The study group included 28 men and 11 women (mean age: 49.1 years, range: 18 to 89 years) (Table 1).

The patients were evaluated in terms of age, gender, cause of death, if the collapse was witnessed, if a bystander performed cardiopulmonary resuscitation (CPR), the initial electrocardiogram (ECG) on admission, the duration of resuscitation, the Glasgow coma scale (GCS) score within 6 h after ROSC, the time between MRI and ROSC, and the Glasgow outcome scale (GOS) score [22]. The resuscitation protocols followed the American Heart Association guidelines [23, 24]. If intracranial haemorrhage was suspected, brain CT was examined as soon as possible after resuscitation. All of the patients were admitted to an intensive care unit (ICU), and they received standard intensive care and monitoring, including mechanical ventilation, arterial catheters, central venous catheters, urinary catheters, and rectal temperature measurements. Neurological examinations were performed at zero, six hours, one day, three days, five days, one week and two weeks after cardiac arrest. SSEPs were performed between one and three days after ROSC. A standardised protocol for therapeutic hypothermia was used in comatose patients during the latter half of the study period. Eligible patients underwent therapeutic hypothermia using an external cooling device for 24 h with a target temperature of 33.0 ± 1°C. Slow rewarming to normal temperature was conducted over eight hours. In patients with therapeutic hypothermia, MRI was performed after normothermia. All of the patients underwent limited MRI that was confined to a DWI and a T2-weighted image (T2WI) for rapid image acquisition (<10 minutes) within five days after resuscitation (the acute and early subacute phases) while avoiding pseudonormalisation of the DWI [15]. The neurological outcomes of the patients were assessed using the GOS score at three months after the cardiac arrest. There was no withdrawal of life support. The comatose patients were divided into two groups: the GOS scores between 1 and 3 (death, vegetative state, and severe disability) were grouped as unfavourable outcomes; and the GOS scores of 4 and 5 (moderate disability and good recovery) were grouped as favourable outcomes. The control group consisted of 16 men and 7 women (mean age: 51.7 years, range: 30 to 80 years) who were examined and scanned at the emergency department for dizziness; they were free of neurological disorders or brain trauma with normal brain MRIs. Informed consent was obtained from the patients' relatives and all controls.

In total, 22 of the 39 patients and 23 control subjects were assessed using a 1.5-T system (Signa Excite; General Electric, Milwaukee, WI, USA) that had echo planar capability. These studies included the following sequences: the axial fast spin-echo T2WIs (4000/1002/2 [TR/TE/NEX] with a 5 mm section thickness) and the axial DWIs (7000/105.2 [TR/TE], a section thickness 5 mm, b values of 0 and 1,000 sec/mm², a field of view 240 × 240, and a matrix size 128 × 128). The other 17 patients were examined with a 1.5-T system (Magnetom Vision Plus; Siemens, Erlangen, Germany) that had echo planar capability, with the following sequences: the axial fast spin-echo T2WIs (4500/99/2 [TR/TE/NEX] with a 5 mm section thickness), and the axial DWIs (5700/139 [TR/TE], a section thickness 5 mm, b values of 0 and 1,000 sec/mm², a field of view 240 × 240, and a matrix size 96 × 128). The ADC maps were automatically generated. Only the DWI data and ADC maps were analysed for this study. The T2WI was used to detect old hyperintense abnormalities to exclude chronic infarction, or it was used as a reference image for this study. The MRIs were reviewed on a standard picture archiving and communication system workstation (Maroview; Marotech, Seoul, Korea). The DWIs together with the ADC maps of all the comatose patients were jointly evaluated by two experienced neuroradiologists blinded to the patients' clinical data. The brain injuries on DWI were categorised into four patterns on the basis of the injury region of the grey matter: normal, isolated cortical injury, isolated deep grey nuclei injury (including caudate nucleus, putamen, and thalamus), and mixed cortical and deep grey nuclei injuries.

The ADC values were only obtained from 22 patients who were examined using the GE Signa Excite due to the use of two kinds of MR scanners. On the workstation, the ADC value of each pixel was constantly displayed on the screen with a movement of a region of interest (ROI) cursor. For each patient, the region of a high signal on the DWI and a low signal on the ADC map was identified. The ROIs were positioned on the areas with a minimum ADC on the ADC maps to produce ADC values for each brain region. If the brain regions were normal, then the ROIs were positioned on the predefined locations (Figure 1). The colour shades were used on the ADC maps to visualise the degree of ADC decrease. Regions of low ADC showed a blue colour; in contrast, regions of high ADC showed a white colour (Figure 2). The colour shades on the ADC maps identified the pixel showing the minimum ADC value in each brain region. The ADC measurements from both sides of the brain were averaged as a patient's ADC value or a control ADC value. ROI sizes varied by region, using 4 mm² for cortex, 10 mm² for the caudate nucleus and putamen and 25 to 40 mm² for the subcortical white matter, thalamus, cerebellum, and pons. The percentage of the patient's ADC, as compared to the average normal control ADC in 15 different brain regions, was computed as a percent ADC value. The person placing the ROIs was blinded to the patient's outcome. To ensure accurate localisation and consistency of the measurements, the ROIs were carefully placed by a single analyst (SPC) who worked in consultation with a neuroradiologist who had 15 years of experience reading MRIs.

The data were expressed as means ± standard deviations. Chi-squared and Fisher's exact tests were used to assess qualitative data (clinical: gender, witnessed arrest, bystander CPR, initial ECG on admission, and cause of arrest; MRI: abnormalities of each region of brain). T-tests were used to compare the quantitative data (clinical: age, resuscitation duration, and time between MRI and ROSC). A one-way ANOVA with the Scheffe post hoc test was applied to study the ADC values in the different regions of the brain. Box and whisker plots were constructed to summarise the distributions of the percent ADC values for the control, the favourable and the unfavourable outcome groups. Spearman's correlation test was used to correlate GOS at three months after ROSC with ADC values of each brain region. Correlations between ADC values of each brain region used a Pearson's correlation. Sensitivity, specificity, positive predictive values (PPV), and negative predictive values (NPV) for predicting unfavourable outcome were calculated using the optimal cutoff values determined by ROC curve analysis. The cutoff level predicting unfavourable outcome with 100% specificity was considered to be optimal. A P value of < 0.05 was considered significant. Statistical analysis was performed using the Statistical Package for Social Sciences (SPSS) version 15.0 (SPSS Inc., Chicago, IL, USA).

During the study period, 240 patients unrelated to trauma suffered out-of-hospital cardiac arrest with attempted resuscitation. Of them, 131 achieved ROSC for more than 20 minutes, and 89 patients were admitted to the hospital alive. Sixty-five patients remained comatose for at least 6 h and survived for more than 24 h after admission. Of the 65 patients, 26 were excluded from this study as follows: lack of MRI data because of early death before MRI (n = 10); MRI delay over five days after ROSC (n = 3); no informed consent (n = 4); cardiac arrest due to intracranial haemorrhage (n = 12); and previous history of Parkinson's disease (n = 1), brain operation (n = 2), or cerebral infarction (n = 1). Thus, 39 were included in this study; 13 patients were assigned to the favourable outcome group (GOS 4 to 5), and 26 were assigned to the unfavourable outcome group (GOS 1 to 3). The mean time to MRI after ROSC for the patients was 52.9 ± 37.5 hours (range, 6 to 119 hours). The clinical characteristics of the patients are summarised in Table 1. There were no significant differences between the two groups, except for the initial ECG rhythm on admission. The mean duration in the intensive care was 11.5 ± 7.6 days (range, 3 to 31 days) in patients with favourable outcome and 21.1 ± 19.6 days (range, 2 to 91 days) in patients with unfavourable outcome. Ten patients (25.6%) died with a mean survival period of 9.2 ± 8.0 days (range, 2 to 29 days). Therapeutic hypothermia was performed in 15 (38%) of 39 analysed patients and 4 of the 15 patients had a favourable outcome. Myoclonic or seizure activities were seen within the first three days after ROSC in 15 patients (38.4%). Of 15 patients, 10 had an unfavourable outcome. Pupillary light reflex was often seen within first three days in patients with both favourable and unfavourable outcome. Eleven patients (28%) who showed loss of pupillary light reflex had an unfavourable outcome. Motor response to pain was absent at three days after cardiac arrest in 15 patients (38%) who had an unfavourable outcome. For 20 (51%) of 39 patients, CT scans were performed within 3 h after the event, and the scans were read as normal for 15 patients. The CT scans of the other five patients were interpreted as having brain edema, and they had an unfavourable outcome. In 20 (51%) of 39 patients, SSEP was examined between one and three days after cardiac arrest. Of them, eight patients who showed no cortical response had an unfavourable outcome.

The cortex and basal ganglia were frequently damaged in the patients but predominantly in the unfavourable outcome group (81% vs. 8%, P < 0.001; and 77% vs. 23%, P = 0.002, respectively). In terms of cortical injuries, the Rolandic (precentral and postcentral), occipital, and parietal cortices had more frequent injury than did the frontal and temporal cortices. The cerebellum and pons had no differences in DWI abnormalities between the favourable and unfavourable outcome groups. The subcortical white matter had no DWI abnormality in any of the patients (Table 2). The neurological outcome in relation to DWI patterns is shown in Table 3. The cortical pattern of injury was seen in one patient (3%), the deep grey nuclei pattern was seen in three patients (8%), the cortex and deep grey nuclei pattern was seen in 21 patients (54%), and normal DWI findings were seen in 14 (36%). There were significant differences in the number of patients with normal findings or mixed cortex and deep grey nuclei injuries between the two groups (Fisher's exact test, P < 0.001). However, the cortical pattern and the deep grey nuclei pattern had no difference in the clinical outcome between the two groups.