Recovery of Apparent Diffusion Coefficient After Ischemia-Induced Spreading Depression Relates to Cerebral Perfusion Gradient
Background and Purpose Transient decreases of the apparent diffusion coefficient (ADC) of water as measured by fast diffusion-weighted imaging (DWI) in the ischemic border zone are thought to reflect cellular swelling associated with spreading depression. DWI and dynamic contrast-enhanced MRI were applied to study the characteristics of spreading depression and the correlation between ADC recovery time and tissue perfusion in focal ischemia.
Methods Serial DWI was performed during remote middle cerebral artery occlusion in rats (n=5) with an echo-planar imaging technique. ADC maps were calculated and ADC values displayed as a function of time in user-defined regions of interest with a time resolution of 12 to 16 seconds. Dynamic contrast-enhanced MRI was performed for qualitative correlation of ADC changes with tissue perfusion.
Results Recovery time of transient ADC decreases correlated with the degree of the perfusion deficit (r=.81, P<.001). Slowly recovering ADC declines were found close to the ischemic core and correlated with severe perfusion deficit, while short-lasting ADC declines were typically found in moderately malperfused or normal tissue. Transient ADC decreases originated in the subcortical and cortical ischemic border zones and propagated along the cortex with a velocity of 2.9±0.9 mm/min.
Conclusions The variation in the recovery time of transient ADC decreases in the ischemic periphery reflects the gradient of the tissue perfusion. Severely delayed recovery time after spreading depression is thought to represent the ischemic penumbra.
- cerebral ischemia, focal
- magnetic resonance imaging
- middle cerebral artery occlusion
- spreading cortical depression
Magnetic resonance imaging is able to detect propagating waves representing cortical SD with high spatial and temporal resolution.1 2 3 4 5 A temporary shrinkage of the extracellular space by approximately 50% occurring during SD6 is believed to be the rationale for the transient decrease in the ADC as measured by DWI.2
SD constitutes a tremendous metabolic stress factor for ischemic tissue and is accompanied by ion shifts, stimulation of glucose metabolism, transiently increased blood flow with subsequent hypoperfusion, and pH decrease.7 8 Neuronal cell death occurs in the ischemic border zone if the additional workload cannot be performed because of low blood flow and reduced energy metabolism.9 10 11
A close correlation between the duration of transient DC shifts in the penumbral zone and the energy state of the tissue seems likely, implying that ischemic tissue with marginal metabolic resources has a long recovery period after SD, while tissue farther away in the periphery more readily copes with the additional demands for substrate delivery.9 12 Until recently, this hypothesis was difficult to prove because of the limited spatial resolution of DC or microelectrode recordings.
The mapping of propagating waves of SD, detected as ADC decreases by echo-planar imaging techniques, overcomes these drawbacks and permits the monitoring of frequency, origin, and propagation of SD-coupled ADC changes on a pixel-by-pixel basis in a multislice mode.5 The combination of serial quantitative ADC measurements with dynamic, contrast-enhanced MR perfusion imaging provides a correlation of ADC recovery time with perfusion deficit in focal ischemia in the rat.
Materials and Methods
Anesthesia and Monitoring
Male Sprague-Dawley rats (weight, 300 to 380 g; n=5) were anesthetized with ketamine (40 mg/kg IM) and xylazine (4 mg/kg IM), tracheotomized, and mechanically ventilated. Anesthesia was maintained with 0.75% to 1.0% halothane. Blood samples for blood gas analysis were obtained through an arterial line, and a femoral catheter was placed for contrast agent injection. Heart rate and Sao2 were continuously monitored by a pulse oximeter. Core temperature was maintained at 37±1.0°C by means of a warm air circulation system and was monitored by a rectal probe. All procedures were approved by the Administrative Panel on Laboratory Animal Care of Stanford University (No. 3700).
Animals were prepared as detailed recently.5 In brief, the external carotid, common carotid, and pterygopalatine arteries were tied off, and an occluding device was inserted into the internal carotid artery. The occluding device consisted of a stiff fishing line (Mono Line, 10 lb; Fitec International) with a 4-0 nylon thread (Ethilon black monofilament nylon 4-0; Ethicon, Inc) glued to its end (Krazy Glue, Borden Inc). The tip of the suture was rounded by heating in an open flame and thickened with glue. A polyethylene tube (PE-50) containing the occluding device was then introduced into the common carotid artery. The tip of the occluding device was positioned at the skull base by advancing it through the PE-50 guide catheter before the animal was fixed in an acrylic plastic (Plexiglas) holder with ear and tooth bars. MCAO was achieved in the magnet by further advancing the tip of the suture 9 to 12 mm into the intracranial internal carotid artery.
MR experiments were performed on a 2.0-T GE CSI system with a spin-echo echo-planar technique (echo time, 88 milliseconds; field of view, 40 mm; slice thickness, 2.5 mm; 64×64-pixel matrix; 1 or 3 slices, 1 average ). Serial DWI covered an observation time of approximately 17 minutes with a repeated sequence of six or eight diffusion gradient amplitudes, with a diffusion-weighting b factor that varied from 0 to 1780 s/mm2. A total of 64 to 85 b value sets were acquired for each slice. Diffusion weighting was applied along the Z direction (long axis of the brain). Each image was acquired in 100 milliseconds with a repetition time of 2 seconds, so that data for each complete ADC calculation were acquired in 12 to 16 seconds for six to eight b values. Perfusion imaging was performed with the same multislice echo-planar technique, but with a repetition time of 1.5 seconds, to follow a bolus injection of 0.3 mmol/kg Gd-DOTMA contrast agent (Nihon Medi-Physics Co Ltd).
The MR protocol began with a high-resolution spin-echo diffusion examination to screen for animals with preocclusion ischemia and was followed by a contrast bolus examination to map the baseline perfusion pattern. The suture was advanced 1 to 2 minutes after the start of the serial diffusion-weighted examination. To correlate the ADC changes in the serial diffusion scan with tissue perfusion, a second perfusion experiment was performed 5 minutes after the dynamic DWI scan, followed by a high-resolution spin-echo diffusion experiment (slice thickness=1 mm approximately 25 minutes after MCAO) to more accurately map the infarcted area.
Data were transferred to an off-line workstation (Sun Microsystems) for postprocessing with the use of customized image display software (MRVision Co). Diffusion images were processed with the use of a linear least-squares fitting algorithm to yield serial ADC maps. The perfusion data were processed to generate maps of relative cerebral blood volume by integrating the ΔR2 transit curves, where ΔR2 is the change in the transverse relaxation rate as defined by
where S0 is the baseline signal intensity, Si is the signal intensity during the first pass of the contrast bolus, and TE is echo time. Additionally, bolus-delay maps were calculated that represent the time from the beginning of the image acquisition to the bolus peak and may be considered a qualitative indicator of tissue perfusion.
With the use of the serial ADC maps, ADC changes over time were analyzed on a pixel-by-pixel basis and displayed as a graph. The ischemic core was defined as those pixels showing an ADC decrease without recovery over the observation time of 17 minutes. The periphery of the ischemic core was analyzed pixel by pixel for similar ADC waveforms (ADC decline and recovery over time). Six to eight contiguous pixels adjacent to the ischemic core displaying similar ADC waveforms were integrated into each ROI. Up to four different ROIs were established from the ischemic core toward the periphery. The number of ROIs that could be discriminated depended on the size of the ischemic core and the peri-infarct tissue. The lesion size was naturally variable, and therefore the number, size, and placement of the ROIs were also variable.
From each ROI, the baseline ADC value before MCAO, maximum ADC decrease, final ADC value, time to onset of ADC decrease from MCAO, and duration from onset to 90% recovery were recorded. The ROIs were transferred to the perfusion data for correlation of ADC waveforms and tissue perfusion (Fig 1⇓).
For the correlation of ADC waveforms with the perfusion data, ROIs were transferred to the perfusion raw data and integrated to signal intensity–time curves. Signal intensity–time curves that did not exhibit a signal intensity drop of at least 10% during bolus passage were considered perfusion loss. The BPT and SID were measured for each ROI. The BPT represents the time from the bolus start in normal control tissue to the bolus peak, and the SID is the maximum decline in signal intensity from baseline (Fig 1⇑). A ratio of ischemic to normal tissue was calculated to correct for the dependence of BPT and SID from the relative nature of the bolus tracking perfusion measurement and to obtain a quantity that we could compare between bolus injections and between animals. The BPT and SID ratios were used as indexes of tissue perfusion.13 14 15 16 In both ratios, 1 indicated normal tissue perfusion and 0 indicated perfusion loss.
All values are presented as mean±SEM. Linear least-squares regression analysis was performed for the correlation of ADC and perfusion ratios. The paired t test was used for comparison of grouped data. Statistical significance was accepted at P<.05.
The physiological variables (Pao2, Paco2, arterial pH, pulse oximetric measurements of Sao2 and heart rate, and body core temperature) remained within the normal range during the entire experiment (Table 1⇓). A summary of the ADC changes is presented in Table 2⇓. All ADC values are expressed in units of 10−3 square millimeters per second. ADC values from ROIs in contralateral control regions were pooled and served as a reference (0.67±0.05).
Nine slices in 5 animals were analyzed pixel by pixel and ROIs formed as described previously. In 4 of 5 animals an area with a permanent ADC decrease could be differentiated from areas with transient ADC decreases and variable recovery times. Pixels close to the ischemic core, as identified by a permanent ADC decrease, had a considerably longer recovery period than those in the ischemic periphery. Fig 2⇓ presents serial ADC maps that show the temporal evolution of the ADC for values below 0.5×10−3 mm2/s. The speed of SD propagation was estimated from the shift of the maximum ADC decrease and the distance of the center of the ROI. The calculated velocity was 2.9±0.9 mm/min over all animals and lesions.
Although pixels with transient ADC decrease were mainly present in cortical areas, they were not exclusively restricted to the cortex but were found in subcortical areas as well. All transient waves originated in the ischemic border zone—either cortical, subcortical, or both—and propagated toward the periphery. Fig 3⇓ shows a coronal slice with a cortical lesion. ADC transients originated in a rim around the ischemic core—subcortical and cortical—and then propagated along the cortex.
In slices with a larger ischemic core area, as in Fig 4⇓, the transient ADC declines were restricted to the cortex. The wave of ADC decrease did not propagate into areas with a permanent ADC decrease or to the contralateral hemisphere. In three slices the wave unequivocally propagated from the MCA territory into normally perfused tissue in the territory of the anterior cerebral artery (Fig 4⇓; ROI 4).
In ROIs with a permanent ADC decrease (ROI 1 in Figs 3⇑ and 4⇑), ADC values decreased from 0.65±0.04 to a maximum of 0.5±0.03 (76.9% of baseline) within 3.5±1.1 minutes after MCAO. ADC values started to decrease as early as 1.3±0.7 minutes after MCAO. The corresponding BPT and SID ratios were 0.1±0.2 and 0.1±0.2, respectively, indicating a severe perfusion deficit. The ROI 1 areas corresponded to the lesion area on the final high-resolution spin-echo DWI.
Adjacent to the ischemic core with a permanent ADC decrease, ROIs with transient ADC decreases and variable recovery times were found. In ROIs with transient changes the overall ADC values started to decrease 2.7±0.8 minutes after MCAO and dropped to 73.8% of the initial baseline values. Fig 5⇓ graphically displays the negative correlation of the ADC recovery time with the BPT and SID ratios (BPT ratio: y=−0.013x+0.78; r=.81; P<.001; SID ratio: y=−0.0144x+0.98; r=77; P<.001). ROIs with a long ADC recovery time were characterized by low BPT and SID ratios, indicating a severe perfusion deficit. The faster the ADC recovered, the higher were the BPT and SID ratios.
ROI 2, representing ADC waveforms with recovery times longer than 21 imaging points (≈5.6 minutes), were present in three slices (2 animals) and were located adjacent to the ischemic core (ROI 1). The latency time between MCAO and start of ADC decrease was 2.2±1.2 minutes. Recovery time of the ADC waveform in those regions was 9.2±0.8 minutes, and the BPT and SID perfusion ratios were 0.4±0.2 and 0.5±0.3, respectively.
ROIs 3 were present in all but one slice and were defined by ADC recovery times of 11 to 20 imaging points (≈2.9 to 5.5 minutes). The latency time was 2.6±0.7 minutes. ADC waveforms lasted an average of 4.2±0.9 minutes and corresponded to BPT and SID perfusion ratios of 0.6±0.1 and 0.7±0.1, respectively.
Still farther away in the periphery, ROIs 4, exhibiting even shorter-lasting, transient ADC changes, were present in all slices (ADC recovery times of 6 to 10 imaging points [≈1.5 to 2.5 minutes]). Latency time was 3.5±0.8 minutes, and ADC waveforms recovered within 2.8±0.2 minutes. The perfusion ratios in these ROIs indicated a moderate perfusion deficit, with BPT and SID ratios of 0.65±0.06 and 0.83±0.08, respectively.
In 3 animals, very short transient ADC declines (ROI 5) were found at the border of the MCA and anterior cerebral artery territories (ADC recovery times of ≤5 imaging points [<1.5 minutes]). Latency time was 3.7±0.6 minutes, and ADC changes recovered within 1.1±0.2 minutes. The corresponding BPT and SID perfusion ratios were 0.75±0.17 and 0.93±0.02, respectively.
Several conclusions may be derived from the results observed: (1) The recovery time of transient ADC waveforms reflects the tissue perfusion deficit, and the ischemic penumbra is thought to be characterized by severely delayed ADC recovery. (2) The latency time between MCAO and ADC transients is short close to the ischemic core and longer in the ischemic periphery, thus reflecting the time required to release triggering factors from the ischemic core and propagate to the observation site. (3) Waves of transient ADC decrease originate in a larger area around the ischemic core, propagate to the periphery with a speed consistent with SD, and are not exclusively restricted to the cortex.
Recovery Time of ADC Transients
The ADC recovery times were linearly correlated with the degree of tissue malperfusion: areas with a permanent ADC decrease (ischemic core) corresponded to a perfusion loss or severe perfusion deficit, whereas areas with a rapid recovery of ADC exhibited normal or slightly reduced tissue perfusion. Between these two extremes, there was a range of different ADC waveforms with correlating perfusion deficits. In general, the shorter the ADC transients, the better was the tissue perfusion. In two animals, very short-lasting ADC transients were found in areas with normal tissue perfusion outside the MCA territory. The ADC recovery time of these waves was approximately 1.1 minutes, thus resembling MRI data from ADC transients after KCl-induced SD.2
Earlier studies reported a linear correlation between blood flow and extracellular K+ clearance in focal ischemia, stressing the need for a sufficient energy supply to restore the membrane potential.17 18 19 In 1986 Nedergaard and Astrup7 were the first to report transient shifts of the DC potential in the infarct rim of focal ischemic rats. They suggested that delayed recovery of the DC potential after SD might indicate metabolic stress in the ischemic rim because normoglycemic rats took a much longer time to recover from depolarization than hyperglycemic rats. They further hypothesized that recovery from SD and hence the duration of the DC shift might depend on the available energy substrates. A comparison of the duration of DC shifts11 or [K+]e potentials12 in the vicinity of focal ischemia and after cortical KCl application revealed shorter-lasting SD events in normal tissue.
Simultaneous laser-Doppler flow measurements close to the position of the DC electrodes showed that in peri-infarct tissue the increased demand of oxygen during SD is not coupled to an adequate increase of blood flow, thus proposing a relationship between SD duration and reduction in cerebral blood flow.10 11
The aforementioned findings support our observations that variable ADC recovery times in the ischemic border zone, a monitor of the reestablishment of membrane potential, were closely correlated with energy substrate deprivation as expressed by the perfusion deficit. If we use the classic definition of the ischemic penumbra as tissue with an intact membrane potential but impaired electric activity,20 it seems reasonable to assume that the penumbra takes a considerably longer time to reestablish ion homeostasis after SD than adjacent tissue farther in the periphery of the ischemic core. This interpretation is supported by a recent study involving DWI and the measurement of biochemical parameters, which indicated that DWI hyperintensity represents the ischemic core and the penumbra.21 While penumbra and ischemic core tissue may appear as a uniform hyperintense area on low-temporal-resolution DWI, the dynamic time-resolved ADC measurements used in our study enabled the discrimination between permanent and transient ADC changes. A recent report provided evidence for the detrimental effect of SD on the penumbral tissue,22 and future follow-up studies of tissue with delayed ADC recovery over time will shed additional light on this issue.
Latency Time Between MCAO and ADC Transients
In our experiments, the latency time between MCAO and ADC transients was short close to the ischemic core and long in the ischemic periphery. Although this finding did not reach statistical significance, the trend was obvious and was observable in each individual experiment. The latency time combines two factors: (1) the time required to induce SD, ie, to release triggering factors from the ischemic core tissue, and (2) the time the propagating wave takes to travel toward the observation area.
In experiments involving K+-sensitive microelectrodes and simultaneous DC recordings from two cortical sites, transient depolarizations have been characterized in the peri-infarct tissue and long-lasting ischemic depolarizations discriminated from episodes of SD-like, short depolarizations (10.7 versus 5.7 minutes).12 The latency time between MCAO and extracellular K+ increase was longer in the SD-like depolarizations (2.5 versus 5.1 minutes), and the authors speculated that this might be due to the time required for the propagating wave of SD to travel from the depolarization source toward the recording electrode. This interpretation is in agreement with our findings of long-lasting ADC waveforms with a short latency time located close to the ischemic core and shorter ADC recovery times with longer latency times in the periphery. One explanation for this relationship is the fact that the latency time in the penumbral tissue where SD originates is predominantly determined by the diffusion time of SD-triggering factors such as potassium and glutamate from the infarct core (factor 1), while in the periphery the propagation time adds to the latency time (factor 2).
Origin and Propagation of SD
Studies of SD in intact rat brain in which multiple electrodes were used revealed that SD is not restricted to the cortex but may penetrate through the amygdala into the caudate-putamen complex and even reenter the neocortex.23 Knowledge about the propagation of SD in focal ischemia is restricted to the brain surface. Multidirectional propagation and origin from different parts of the peri-infarct tissue were reported by investigators using two DC electrodes,10 but little is known about propagation of SD outside the ischemic vascular territory or the origin of SD in the peri-infarct tissue and in subcortical areas.
Our findings support the predominant occurrence of SD in cortical layers. However, in slices with small cortical lesions we observed a rim expanding around the ischemic core. This rim seemed to resemble the origin of the propagating wave of ADC decrease. No entry from the cortex to subcortical areas or the contralateral hemisphere was noted. If we assume that SD is not a phenomenon that is limited to neurons but affects glial cells as well,24 25 the origin of SD propagation in subcortical areas seems plausible. In six of nine slices the transient ADC decrease propagated toward the midline, whereas in three slices a bidirectional propagation was observed, with ADC waves traveling toward the base of the brain and the midline. Since we applied multislice measurements in only two animals, no definite conclusions can be drawn regarding the three-dimensional propagation, but previous studies support the view of a concentric planar propagation in both axes.1 2 4 We are presently optimizing the applied MRI sequences to cover the entire ischemic volume with contiguous thin slices.
Combining serial ADC measurements with perfusion imaging, we showed that SD propagates to remote areas outside the affected arterial territory. Few reports provide evidence for ischemia-related SD in normal perfused tissue, since most measurements of cerebral blood flow were concentrated in the peri-infarct tissue. Dietrich et al26 recorded SD in brain areas remote from photothrombotic infarcted regions and pointed out that remote SD might be the underlying reason for induction of immediate early genes throughout the ipsilateral hemisphere after focal infarction.27 Serial ADC measurements provide the basis for further studies on this topic.
In summary, we have shown the following: (1) Transient ADC declines can be detected in the border zone of the ischemic core that are related to SD and that propagate into normally perfused areas. (2) The recovery time of transient ADC decreases in the ischemic periphery reflects the gradient of the tissue perfusion deficit from the ischemic core with no ADC recovery and complete perfusion loss, to severely delayed recovery with severe perfusion deficit, to progressively shorter-lasting ADC waveforms and subsequently better tissue perfusion. We hypothesize that delayed ADC waveform recovery reflects the infarct penumbra. (3) The latency time of transient ADC decreases after MCAO is shorter the closer the observed area is to the ischemic core, thus reflecting the time necessary to trigger SD and the propagation time. (4) Additionally, we have shown that SD can originate in a rim around the ischemic core that involves cortical and subcortical tissues.
Selected Abbreviations and Acronyms
|ADC||=||apparent diffusion coefficient|
|DWI||=||diffusion-weighted magnetic resonance imaging|
|MCA||=||middle cerebral artery|
|MCAO||=||middle cerebral artery occlusion|
|ROI(s)||=||region(s) of interest|
|Sao2||=||saturation with oxygen, arterial blood|
|SID||=||signal intensity decline|
This study was supported by a grant from the Deutsche Forschungsgemeinschaft (Dr Röther).
- Received October 10, 1995.
- Revision received January 18, 1996.
- Accepted January 18, 1996.
- Copyright © 1996 by American Heart Association
Hansen AJ, Olsen CE. Brain extracellular space during spreading depression and ischemia. Acta Physiol Scand. 1980;355-365.
Minematsu K, Fisher M, Li L, Sotak CH. Diffusion and perfusion magnetic resonance imaging studies to evaluate noncompetitive N-methyl-d-aspartate antagonist and reperfusion in experimental stroke in rats. Stroke. 1993;24:2074-2081.
Astrup J, Symon L, Branston NM, Lassen NA. Cortical evoked potential and extracellular K+ and H+ at critical levels of brain ischemia. Stroke. 1977;88:51-57.
Astrup J, Siesjo BK, Symon L. Thresholds in cerebral ischemia: the ischemic penumbra. Stroke. 1981;12:723-725.
Kohno K, Hoehn-Berlage M, Mies G, Back T, Hossmann KA. Relationship between diffusion-weighted magnetic resonance images, cerebral blood flow, and energy state in experimental brain infarction. Magn Reson Med. 1995;13:65-71.
Busch E, Gyngell M, Eis M, Hoehn-Berlage M, Hossmann KA. Spreading depression-induced cerebral infarct growth observed by diffusion-weighted MRI. In: Program and abstracts of the third annual meeting of the Society of Magnetic Resonance; August 18-25, 1995; Nice, France. Abstract, p 26.
Kraig PR, Dong L, Thisted R, Jaeger CB. Spreading depression increases immunohistochemical staining of glial fibrillary acidic protein. J Neurosci.. 1991;11:2187-2198.