1180
Clinical neuroscience
Timing the ischaemic stroke by 1H-MRI: improved accuracy using absolute relaxation times over signal intensities Harriet J. Rogersa,*, Bryony L. McGarrya,*, Michael J. Knighta, Kimmo T. Jokivarsib, Olli H.J. Gröhnb and Risto A. Kauppinena One in four ischaemic stroke patients are ineligible for thrombolytic treatment due to unknown onset time. Quantification of absolute MR relaxation times and signal intensities are potential methods for estimating stroke duration. We compared the accuracy of these approaches and determined whether changes in relaxation times and signal intensities identify the same ischaemic tissue as diffusion MRI. Seven Wistar rats underwent permanent middle cerebral artery occlusion to induce focal ischaemia and were scanned at six time points. The trace of the diffusion tensor (DAV), T1ρ and T2 were acquired at 4.7 T. Results show relaxation times, and signal intensities of the MR relaxation parameters increase linearly with ischaemia duration (P < 0.001). Using T1ρ and T2 relaxation times, an estimate of 4.5 h after occlusion has an uncertainty of ± 12 and ± 35 min, respectively, compared with over 50 min for signal intensities. In addition, we present a pixel-by-pixel method that simultaneously estimates stroke onset time and identifies potentially irreversible ischaemic tissue using absolute relaxation times. This method demonstrates signal intensity changes during ischaemia display an ambiguous
Introduction The viability of brain tissue decreases and the lesion volume increases over time in the ischaemic brain. Hence, objective means of evaluating brain tissue status are actively sought in clinical environments. Stroke onset time is a key factor in all current treatment procedures [1]. For instance, patients are almost always deemed ineligible for thrombolytic therapy if onset time is unknown, stroke duration is greater than 4.5 h or there is an imminent risk of haemorrhage [1]. A preclinical study using 23Na-MRI reported a linear increase of total brain sodium with time from stroke onset [2]. Although this suggests that 23Na-MRI could be used to estimate stroke duration [2], this approach requires dedicated hardware that is not commonplace in clinical MR systems. Studies using 1H-MRI have shown that absolute relaxation times Τ1ρ (T1 relaxation in the rotating frame) and T2 are informative of brain tissue viability in the early moments of ischaemia [3,4]. Differences in T2 and Τ1ρ between homologous regions of ischaemic and contralateral nonischaemic hemispheres (ΔT2, ΔΤ1ρ) increase linearly with time from stroke onset and may therefore serve as proxies for stroke duration [5–7]. A clinical MRI study reported quadratic dependency of 0959-4965 © 2014 Wolters Kluwer Health | Lippincott Williams & Wilkins
pattern and highlights the possibility that diffusion MRI overestimates the true extent of irreversible ischaemia. In conclusion, quantification of absolute relaxation times at a single time point enables a more accurate estimation of stroke duration than signal intensities and provides more information about tissue status in ischaemia. NeuroReport 25:1180–1185 © 2014 Wolters Kluwer Health | Lippincott Williams & Wilkins. NeuroReport 2014, 25:1180–1185 Keywords: irreversible ischaemia, ischaemic stroke, magnetic resonance imaging, onset time, relaxation times, signal intensities, T2, T1ρ a School of Experimental Psychology and Bristol Clinical Research and Imaging Centre, University of Bristol, Bristol, UK and bDepartment of Neurobiology, A.I. Virtanen Institute, University of Eastern Finland, Kuopio, Finland
Correspondence to Risto A. Kauppinen, MD, PhD, School of Experimental Psychology, University of Bristol, 12a Priory Road, Clifton, Bristol BS8 1TU, UK Tel: + 44 1179288461; e-mail: [email protected] *Harriet J. Rogers and Bryony L. McGarry contributed equally to the writing of this article. Received 6 May 2014 accepted 2 July 2014
ΔT2 and ischaemia duration, enabling estimation of onset time with 79% accuracy [6]. The apparent diffusion coefficient (ADC) in the brain precipitously decreases in early moments of ischaemia by up to 60% of its original value [8]. While ADC may be suggestive of tissue viability [9], preclinical evidence suggests that ADC is uninformative of ischaemia duration in the first 12 h [4]. Indeed, ΑDC has little utility in estimating onset time of acute ischaemic stroke patients [10]. A combination of diffusion-weighted imaging (DWI) and T2-Fluid Attenuated Inversion Recovery (FLAIR) MRI is an additional method that has been proposed for stroke duration estimates [11]. Patients with a positive DWI lesion and unchanged T2-FLAIR signal are assumed to be within the 4.5 h treatment-window [11]. This procedure effectively exploits the time-dependent T2 increase in ischaemia but has low sensitivity causing misclassification of patients [11,12] and has been criticized for its reliance on visual assessment [12,13]. To overcome these issues quantification of interhemispheric T2FLAIR signal intensities has been suggested [13,14]. Although T2-FLAIR signal intensities have been found to have a moderate positive correlation with stroke DOI: 10.1097/WNR.0000000000000238
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Timing the ischaemic stroke by 1H-MRI Rogers et al. 1181
duration in canine stroke models [14] and patient scans [13], the viability of this approach remains unclear [13].
Each pixel in the resulting image represented the APD between homologous pixels in both hemispheres.
As both signal intensities and absolute relaxation times have been suggested as potential methods for estimating stroke duration [5,6,13,14], this study first aimed to compare the accuracy of these methods for ΔΤ1ρ and ΔT2 using a rat model of permanent focal ischaemia. The second goal was to determine whether significant differences in these parameters identify the same ischaemic tissue as diffusion MRI and whether this could serve as another proxy for onset time.
Confidence interval maps were computed from APD images to determine whether hemispheric differences were statistically significant. To do this, pixel-wise one-sample t-tests were performed, where the uncertainty for each pixel was estimated by calculating the SD of intensities or relaxation times over a 7 × 7 pixel kernel surrounding the pixel to be tested. This method assumed the image to be locally smooth. Confidence intervals of the t-statistic of each pixel were calculated to generate the final maps.
Materials and methods
Data analysis
Animal model
To locate ischaemic tissue, visually guided regions of interest (ROIs) were manually drawn around the clear visual boundaries of hypointense regions of the striatum on DAV images. This approach was used to focus on the infarct core (the first tissue to become ischaemic), as in clinical settings the timing is determined from the first ischaemic event [1]. ROIs were drawn for each rat at each time point. DAV defined ROIs were loaded onto APD images to extract mean differences between hemispheres for signal intensities and relaxation times (ΔT1ρ, ΔT2). To determine the relationship between time since MCAO and ΔT1ρ or ΔT2, Pearson’s correlations were calculated. One and two parameter linear models were fitted to the data to estimate stroke duration. Selection of models was determined by F-tests returning a P value of less than 0.05 and the size of the error for an estimate of 4.5 h. In line with linear least squares assumptions the uncertainty associated with estimates of stroke duration were predicted from the fitted parameters and their associated confidence intervals according to the following formula [15]: vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi uX u q ! 2 2 t Dt ¼ f p Dpj ; qpj j
Male Wistar rats (n = 7) underwent permanent middle cerebral artery occlusion (MCAO) to induce focal stroke [5]. Animals were euthanized by intravenous injection of saturated KCl at the end of the experiment. Animal procedures were conducted according to European Community Council Directives 86/609/EEC guidelines and approved by the Animal Care and Use Committee of the University of Eastern Finland. Magnetic resonance imaging
MRI data were acquired according to methods described by Jokivarsi et al. [5]. The trace of diffusion tensor (DAV= 1/3 trace [D]) images, Τ1ρ and T2 were obtained at 30–60 min intervals at 4.7 T for 6 h after induction of focal brain ischaemia. Five echo times (10, 20, 40, 60 and 80 ms) and five spin-lock pulse durations (10, 20, 40, 60 and 80 ms) were used for T2 and Τ1ρ MRI acquisitions, respectively. Total acquisition time for DAV, T2 and Τ1ρ was 24 min. Image computation
Image postprocessing was performed using the MRI data analysis software Mango (Research Imaging Institute, UT Health Science Center at San Antonio, Texas, USA) and software written in-house using the Matlab programming environment (MathWorks, Natick, Massachusetts, USA). Images used for the T1ρ and T2 signal intensity analyses were the sum of weighted images acquired at each echo time (T2) and spin-lock duration (T1ρ). T1ρ and T2 relaxation maps were computed using a monoexponential fit in a logarithmic space. To eliminate inter-subject variation, changes in relaxation times and absolute signal intensities were quantified relative to the nonischaemic hemisphere. Thereby each rat served as its own control. To compare ischaemic and nonischaemic hemispheres absolute percentage difference (APD) images were computed for weighted images and relaxometry maps for each rat and time point. Where entire images were mirrored about the vertical axis and manually coregistered to their originals, followed by the calculation: APD = 100 × [(original − mirrored)/(original + mirrored)].
! where f p is the model as function of fitted parameters ! p and Δpj is the confidence interval for parameter j. DAV defined ROIs were loaded onto confidence interval maps. Pixels within the ROI were colour-coded according to their confidence interval. The per cent of pixels with confidence intervals more than 95% within the ROI were calculated. To determine the relationship between stroke duration and the per cent of significant pixels within the ROI, Pearson’s correlations were conducted. The above calculation for estimating uncertainty associated with estimates of stroke duration was used.
Results ΔT1ρ or ΔT2 relaxation times and signal intensities show a significant positive correlation with stroke duration allowing stroke onset to be estimated (Fig. 1). Relaxation times provide a substantially more accurate estimate of stroke duration
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
1182 NeuroReport 2014, Vol 25 No 15
compared to signal intensities. For example, using relaxation times an estimate of 4.5 h has an uncertainty of ± 12 min for ΔT1ρ, ± 35 min for ΔT2, compared with ± 59 and ± 53 min using respective ΔT1ρ and ΔT2 signal intensities.
The amount of pixels with significant changes in T1ρ and T2 within DAV defined ROIs increases with time (Fig. 2). Both increases and decreases are seen in T2 relaxation times (Fig. 2c), but not in T1ρ. It is also evident that
Fig. 1
(a)
(b) 500
400
400
Time post-MCAO (min)
300 300
200
200
100 100 0
0
100 0
2
6 4 8 ΔT1ρ absolute relaxation times
10
12
Time post-MCAO (min)
(c) 400
(d) 400
300
300
200
200
100
100
0
−4
−2
0 6 0 2 4 ΔT2 absolute relaxation times
8
Individual rats
10
−2
−4
0
−2
1σ error bars
2 4 6 ΔT1ρ signal intensities
0
2 4 6 ΔT2 signal intensities
8
8
10
10
Linear trend line
ΔT1ρ and ΔT2 relaxation times and signal intensities as a function of ischaemia duration and linear trend lines which estimated time since middle cerebral artery occlusion (MCAO). Data are averaged over the regions of interest for each rat at each time point and the fitting performed for the pooled data over all rats. (a) ΔT1ρ relaxation times, MCAO = 29.9 (± 1.4) × ΔT1ρ, r = 0.96, P < 0.001. (b) ΔT1ρ signal intensities, MCAO = 30.3 (± 7.3) × ΔT1ρ + 69.4 (± 33.4), r = 0.81, P < 0.001. (c) ΔT2 relaxation times, MCAO = 24.1 (± 3.7) × ΔT2 + 86.7 (± 20.5), r = 0.90, P < 0.001. (d) ΔT2 signal intensities, MCAO = 22.2 (± 6.5) × ΔT2 + 117.3 (± 29.0), r = 0.75, P < 0.001.
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Timing the ischaemic stroke by 1H-MRI Rogers et al. 1183
Fig. 2
(a) Time post-MCAO ± 4 min
DAV
(b)
(c)
T1 Signal intensities
Relaxometry maps
T2 Signal intensities
Relaxometry maps
1h
2h
3h
4h
5h
6h
Confidence intervals: 99 - 100% 95 - 90% 90 - 95%
Negative difference Positive difference
≤ 90% DAV and confidence interval maps of one representative rat during the progression of ischaemia. Confidence interval maps are overlaid with weighted images and relaxometry maps for clarity. (a) DAV images used to define the region of interest (ROI). (b) Significant ΔT1ρ within DAV defined ROIs for signal intensities and relaxation times. (c) Significant ΔT2 within the DAV defined ROIs for signal intensities and relaxation times, the latter showing positive and negative ΔT2.
signal intensities demonstrate an ambiguous effect of ischaemia over time making it unsuitable for estimating stroke onset. This was observed for all animals. For both T1ρ and T2 the per cent of significant pixels (P < 0.05) within DAV defined ROIs shows a significant positive correlation with time post-MCAO allowing onset time to be estimated (Fig. 3). For example, an estimate of 4.5 h has an uncertainty of ± 19 min for T1ρ and ± 26 min for T2. To assess image quality, signal-to-noise ratio SNR : mcontralateral signal= mcontralateral noise and contrast-to mischaemic signal mcontralateral signal p ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi noise ratio CNR : were calcu2 2 nischaemic noise þncontralateral noise
lated across all rats at 4.5 h and local noise was estimated using a 5 × 5 kernel. SNR for relaxometry maps (T1ρ: 19.1, SD = 2.2; T2: 17.5, SD = 2.2) and summedweighted images (T1ρ: 18.3, SD = 1.2; T2: 17.5, SD = 1.2) were comparable, demonstrating similar SNR
for weighted images and relaxometry maps. CNR values for relaxometry maps (T1ρ: 2.8, SD = 0.5; T2: 2.0, SD = 0.7) were greater than for weighted images (T1ρ: 0.7, SD = 0.9; T2: 0.4, SD = 1.1). Average relaxation times from an ROI placed in the striatum of the nonischaemic hemisphere of scans acquired at the first time point were: T2 = 56.5 ms (SD = 1.7) and T1ρ = 74.3 ms (SD = 2.3).
Discussion The present study shows that quantification of the absolute difference in relaxation times between the ischaemic and contralateral hemisphere provides a more accurate estimate of ischaemia duration compared with respective signal intensity differences. We show that the margin of error associated with relaxation time estimates is far smaller than when using signal intensities. Issues using signal intensities for assessment of brain tissue status have been documented [13,16]. Kavec et al. [16] reported that following 30 min MCAO in rats,
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
1184
NeuroReport 2014, Vol 25 No 15
Fig. 3
(b)
(a) 400
400
Time post-MCAO (min)
300 300
200 200
100
100
0
0 0
20
40
60
80
Per cent of DAV ROI with significant ΔT1
100
0
20
40
60
80
100
Per cent of DAV ROI with significant ΔT2
The per cent of pixels within the DAV defined region of interest (ROI) with significant ΔT1ρ and ΔT2 relaxation times as a function of ischaemia duration. (a) MCAO = 3.6(± 0.3) × ΔT1ρ, r = 0.89, P < 0.001. (b) MCAO = 4.6(± 0.4) × ΔT2, r = 0.79, P < 0.001. MCAO, middle cerebral artery occlusion.
ischaemic damage was not evident at 24 h in weighted images but could be identified with absolute T1ρ and T2. Furthermore, quantification of T2-FLAIR signal intensities were found to have low sensitivity leading patients within the treatment-window to be misclassified [13]. In addition to higher CNR, relaxation times may provide a more accurate estimate of stroke onset as inherent variation due to B0 and B1 inhomogeneities are eliminated by fitting signal intensities to the monoexponential relaxation equation. One could potentially improve contrast due to ischaemia in weighted images by optimizing either the echo or spin-lock time; however, to offset the worsened SNR in required long TE or spin-lock time images, scan times may increase. A shortening of T2 was observed during the initial stages of ischaemia (Fig. 2c), an effect presumably dominated by the negative blood–oxygenation level-dependent (BOLD) effect [3,17]. At a certain time point T2 therefore, displays no difference to the contralateral hemisphere and could confound estimates of ischaemia duration. This however, occurs in the first 2 h of the insult when ischaemia may be reversible. Thomalla et al. [11] found patients without T2-FLAIR hyperintensity in DWI defined ischaemic tissue had stroke duration of less than 3 h. Thus, patients showing no significant difference in T2 relaxation times due to T2 cross-over may be within the recanalization treatment-window. T1ρ does not
shorten during ischaemia due to low sensitivity of the BOLD effect [18] and present results show it provides a more accurate estimate of stroke duration than ΔT2. Nevertheless, T1ρ is not widely used in clinical neuroMRI due to associated SAR issues. Recent advancements in MRI pulse sequencing suggest that through use of adiabatic T1ρ, imaging of the brain with good volume coverage is feasible with greatly reduced SAR both for preclinical [19] and clinical applications [20]. It has been suggested that patients with hypoperfused tissue may still benefit from thrombolysis [21]. Severe ADC decreases are typically assumed to represent irreversible tissue damage, for instance, 20% decrease in ADC of 7 h of permanent MCAO has been observed to result in tissue damage as revealed by histology [7]. However, clinical data suggest that severely reduced ADC in acute phase overestimates infarct volume [22]. Hohn-Berlage et al. [7] reported that in the early moments of ischaemia the area of elevated T2 relaxation time is smaller than the area of reduced ADC. We extended this observation using statistical levels of significance and found that the DAV defined ROIs contain pixels with normal appearing T1ρ and T2 relaxation. These may or may not represent potentially reversible ischaemic tissue. The discrepancy between DAV and T1ρ and T2 potentially indicates transition to irreversible ischaemia takes place with time. In addition, the linear
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Timing the ischaemic stroke by 1H-MRI Rogers et al. 1185
relationship between ischaemia duration and the per cent of pixels with significant ΔΤ1ρ and ΔΤ2 in the DAV ROI provides a further estimate of onset time. In this instance it should be borne in mind that each MRI pixel contains a number of neural cell types that vary regionally and have inherently different sensitivity to ischaemia. Indeed, recent data from rodent stroke models point to variation in MR relaxation time responses both in size and kinetics during ischaemia within low ADC volume [7,19].
8
9
10
11
Conclusion The present data demonstrate that absolute relaxation times are more accurate than signal intensities for MRI assessment of ischaemic stroke, both in estimating onset time and identifying ischaemic tissue. Given this, and that acquisition time would be reasonable in clinical settings (< 15 min), we encourage the use of absolute relaxation times in assessment of acute ischaemic stroke.
12
13
14
Acknowledgements This study was supported by grants from the Academy of Finland and the Sigrid Juselius Foundation.
15 16
Conflicts of interest
There are no conflicts of interest.
References 1 Hacke W, Kaste M, Bluhmki E, Brozman M, Dávalos A, Guidetti D, et al. Thrombolysis with alteplase 3 to 4.5 h after acute ischemic stroke. N Engl J Med 2008; 359:1317–1329. 2 Jones SC, Kharlamov A, Yanovski B, Kim DK, Easley KA, Yushmanov VE, et al. Stroke onset time using sodium MRI in rat focal cerebral ischemia. Stroke 2006; 37:883–888. 3 Gröhn OHJ, Kettunen MI, Mäkelä HI, Penttonen M, Pitkänen A, Lukkarinen JA, Kauppinen RA. Early detection of irreversible cerebral ischemia in the rat using dispersion of the magnetic resonance imaging relaxation time, T1p. J Cereb Blood Flow Metab 2000; 20:1457–1466. 4 Knight RA, Dereski MO, Helpern JA, Ordidge RJ, Chopp M. Magnetic resonance imaging assessment of evolving focal cerebral ischemia. Comparison with histopathology in rats. Stroke 1994; 25:1252–1261. 5 Jokivarsi KT, Hiltunen Y, Gröhn H, Tuunanen P, Gröhn OHJ, Kauppinen RA. Estimation of the onset time of cerebral ischemia using T1p and T2 MRI in rats. Stroke 2010; 41:2335–2340. 6 Siemonsen S, Mouridsen K, Holst B, Ries T, Finsterbusch J, Thomalla G, et al. Quantitative T2 values predict time from symptom onset in acute stroke patients. Stroke 2009; 40:1612–1616. 7 Hoehn-Berlage M, Eis M, Back T, Kohno K, Yamashita K. Changes of relaxation times (T1, T2) and apparent diffusion coefficient after permanent middle cerebral artery occlusion in the rat: temporal evolution, regional extent, and comparison with histology. Magn Reson Med 1995; 34:824–834.
17
18
19
20
21
22
Hoehn-Berlage M, Norris D, Kohno K, Mies G, Leibfritz D, Hossmann K. Evolution of regional changes in apparent diffusion coefficient during focal ischemia of rat brain: the relationship of quantitative diffusion NMR imaging to reduction in cerebral blood flow and metabolic disturbances. J Cereb Blood Flow Metab 1995; 15:1002–1011. Hasegawa Y, Fisher M, Latour LL, Dardzinski BJ, Sotak CH. MRI diffusion mapping of reversible and irreversible ischemic injury in focal brain ischemia. Neurology 1994; 44:1484–14890. Madai V, Galinovic I, Grittner U, Zaro-Weber O, Schneider A, Martin S, et al. DWI intensity values predict FLAIR lesions in acute ischaemic stroke. PLoS One 2014; 9:e92295. Thomalla G, Cheng B, Ebinger M, Hao Q, Tourdias T, Wu O, et al. DWIFLAIR mismatch for the identification of patients with acute ischaemic stroke within 4·5 h of symptom onset (PRE-FLAIR): a multicentre observational study. Lancet Neurol 2011; 10:978–986. Ebinger M, Galinovic I, Rozanski M, Brunecker P, Endres M, Fiebach JB. Fluid-attenuated inversion recovery evolution within 12 h from stroke onset: a reliable tissue clock? Stroke 2010; 41:250–255. Cheng B, Brinkmann M, Forkert ND, Treszl A, Ebinger M, Köhrmann M, et al. STIR and VISTA Imaging Investigators. Quantitative measurements of relative fluid-attenuated inversion recovery (FLAIR) signal intensities in acute stroke for the prediction of time from symptom onset. J Cereb Blood Flow Metab 2013; 33:76–84. Xu X, Zu Q, Lu S, Cheng Q, Yu J, Sheng Y, et al. Use of FLAIR imaging to identify onset time of cerebral ischaemia in a canine model. AJNR Am J Neuroradiol 2014; 35:311–316. Ku H. Notes on the use of propagation of error formulas. J Res Natl Bur Stand (1934) 1966; 70:263–273. Kavec M, Gröhn OHJ, Kettunen MI, Silvennoinen MJ, Garwood M, Kauppinen RA. Acute cerebral ischemia in rats studied by Carr-Purcell spinecho magnetic resonance imaging: assessment of blood oxygenation leveldependent and tissue effects on the transverse relaxation. Magn Reson Med 2004; 51:1138–1146. Calamante F, Lythgoe MF, Pell GS, Thomas DL, King MD, Busza AL, et al. Early changes in water diffusion, perfusion, T1, and T2 during focal cerebral ischemia in the rat studied at 8.5 T. Magn Reson Med 1999; 4:479–485. Kettunen MI, Gröhn OH, Penttonen M, Kauppinen RA. Cerebral T1rho relaxation time increases immediately upon global ischemia in the rat independently of blood glucose and anoxic depolarization. Magn Reson Med 2001; 46:565–572. Jokivarsi KT, Niskanen J-P, Michaeli S, Gröhn HI, Garwood M, Kauppinen RA, Gröhn OH. Quantitative assessment of water pools by T1ρ and T2 MRI in acute cerebral ischemia of the rat. J Cereb Blood Flow Metab 2009; 29:206–216. Andronesi OC, Bhat H, Reuter M, Mukherjee S, Caravan P, Rosen BR. Whole brain mapping of water pools and molecular dynamics with rotating frame MR relaxation using gradient modulated low-power adiabatic pulses. Neuroimage 2014; 89:92–109. Albers GW, Thijs VN, Wechsler L, Kemp S, Schlaug G, Skalabrin E, et al. DEFUSE Investigators. Magnetic resonance imaging profiles predict clinical response to early reperfusion: the diffusion and perfusion imaging evaluation for understanding stroke evolution (DEFUSE) study. Ann Neurol 2006; 60:508–517. Fiehler J, Foth M, Kucinski T, Knab R, von Bezold M, Weiller C, et al. Severe ADC decreases do not predict irreversible tissue damage in humans. Stroke 2002; 33:79–86.
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Clinical neuroscience
Timing the ischaemic stroke by 1H-MRI: improved accuracy using absolute relaxation times over signal intensities Harriet J. Rogersa,*, Bryony L. McGarrya,*, Michael J. Knighta, Kimmo T. Jokivarsib, Olli H.J. Gröhnb and Risto A. Kauppinena One in four ischaemic stroke patients are ineligible for thrombolytic treatment due to unknown onset time. Quantification of absolute MR relaxation times and signal intensities are potential methods for estimating stroke duration. We compared the accuracy of these approaches and determined whether changes in relaxation times and signal intensities identify the same ischaemic tissue as diffusion MRI. Seven Wistar rats underwent permanent middle cerebral artery occlusion to induce focal ischaemia and were scanned at six time points. The trace of the diffusion tensor (DAV), T1ρ and T2 were acquired at 4.7 T. Results show relaxation times, and signal intensities of the MR relaxation parameters increase linearly with ischaemia duration (P < 0.001). Using T1ρ and T2 relaxation times, an estimate of 4.5 h after occlusion has an uncertainty of ± 12 and ± 35 min, respectively, compared with over 50 min for signal intensities. In addition, we present a pixel-by-pixel method that simultaneously estimates stroke onset time and identifies potentially irreversible ischaemic tissue using absolute relaxation times. This method demonstrates signal intensity changes during ischaemia display an ambiguous
Introduction The viability of brain tissue decreases and the lesion volume increases over time in the ischaemic brain. Hence, objective means of evaluating brain tissue status are actively sought in clinical environments. Stroke onset time is a key factor in all current treatment procedures [1]. For instance, patients are almost always deemed ineligible for thrombolytic therapy if onset time is unknown, stroke duration is greater than 4.5 h or there is an imminent risk of haemorrhage [1]. A preclinical study using 23Na-MRI reported a linear increase of total brain sodium with time from stroke onset [2]. Although this suggests that 23Na-MRI could be used to estimate stroke duration [2], this approach requires dedicated hardware that is not commonplace in clinical MR systems. Studies using 1H-MRI have shown that absolute relaxation times Τ1ρ (T1 relaxation in the rotating frame) and T2 are informative of brain tissue viability in the early moments of ischaemia [3,4]. Differences in T2 and Τ1ρ between homologous regions of ischaemic and contralateral nonischaemic hemispheres (ΔT2, ΔΤ1ρ) increase linearly with time from stroke onset and may therefore serve as proxies for stroke duration [5–7]. A clinical MRI study reported quadratic dependency of 0959-4965 © 2014 Wolters Kluwer Health | Lippincott Williams & Wilkins
pattern and highlights the possibility that diffusion MRI overestimates the true extent of irreversible ischaemia. In conclusion, quantification of absolute relaxation times at a single time point enables a more accurate estimation of stroke duration than signal intensities and provides more information about tissue status in ischaemia. NeuroReport 25:1180–1185 © 2014 Wolters Kluwer Health | Lippincott Williams & Wilkins. NeuroReport 2014, 25:1180–1185 Keywords: irreversible ischaemia, ischaemic stroke, magnetic resonance imaging, onset time, relaxation times, signal intensities, T2, T1ρ a School of Experimental Psychology and Bristol Clinical Research and Imaging Centre, University of Bristol, Bristol, UK and bDepartment of Neurobiology, A.I. Virtanen Institute, University of Eastern Finland, Kuopio, Finland
Correspondence to Risto A. Kauppinen, MD, PhD, School of Experimental Psychology, University of Bristol, 12a Priory Road, Clifton, Bristol BS8 1TU, UK Tel: + 44 1179288461; e-mail: [email protected] *Harriet J. Rogers and Bryony L. McGarry contributed equally to the writing of this article. Received 6 May 2014 accepted 2 July 2014
ΔT2 and ischaemia duration, enabling estimation of onset time with 79% accuracy [6]. The apparent diffusion coefficient (ADC) in the brain precipitously decreases in early moments of ischaemia by up to 60% of its original value [8]. While ADC may be suggestive of tissue viability [9], preclinical evidence suggests that ADC is uninformative of ischaemia duration in the first 12 h [4]. Indeed, ΑDC has little utility in estimating onset time of acute ischaemic stroke patients [10]. A combination of diffusion-weighted imaging (DWI) and T2-Fluid Attenuated Inversion Recovery (FLAIR) MRI is an additional method that has been proposed for stroke duration estimates [11]. Patients with a positive DWI lesion and unchanged T2-FLAIR signal are assumed to be within the 4.5 h treatment-window [11]. This procedure effectively exploits the time-dependent T2 increase in ischaemia but has low sensitivity causing misclassification of patients [11,12] and has been criticized for its reliance on visual assessment [12,13]. To overcome these issues quantification of interhemispheric T2FLAIR signal intensities has been suggested [13,14]. Although T2-FLAIR signal intensities have been found to have a moderate positive correlation with stroke DOI: 10.1097/WNR.0000000000000238
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Timing the ischaemic stroke by 1H-MRI Rogers et al. 1181
duration in canine stroke models [14] and patient scans [13], the viability of this approach remains unclear [13].
Each pixel in the resulting image represented the APD between homologous pixels in both hemispheres.
As both signal intensities and absolute relaxation times have been suggested as potential methods for estimating stroke duration [5,6,13,14], this study first aimed to compare the accuracy of these methods for ΔΤ1ρ and ΔT2 using a rat model of permanent focal ischaemia. The second goal was to determine whether significant differences in these parameters identify the same ischaemic tissue as diffusion MRI and whether this could serve as another proxy for onset time.
Confidence interval maps were computed from APD images to determine whether hemispheric differences were statistically significant. To do this, pixel-wise one-sample t-tests were performed, where the uncertainty for each pixel was estimated by calculating the SD of intensities or relaxation times over a 7 × 7 pixel kernel surrounding the pixel to be tested. This method assumed the image to be locally smooth. Confidence intervals of the t-statistic of each pixel were calculated to generate the final maps.
Materials and methods
Data analysis
Animal model
To locate ischaemic tissue, visually guided regions of interest (ROIs) were manually drawn around the clear visual boundaries of hypointense regions of the striatum on DAV images. This approach was used to focus on the infarct core (the first tissue to become ischaemic), as in clinical settings the timing is determined from the first ischaemic event [1]. ROIs were drawn for each rat at each time point. DAV defined ROIs were loaded onto APD images to extract mean differences between hemispheres for signal intensities and relaxation times (ΔT1ρ, ΔT2). To determine the relationship between time since MCAO and ΔT1ρ or ΔT2, Pearson’s correlations were calculated. One and two parameter linear models were fitted to the data to estimate stroke duration. Selection of models was determined by F-tests returning a P value of less than 0.05 and the size of the error for an estimate of 4.5 h. In line with linear least squares assumptions the uncertainty associated with estimates of stroke duration were predicted from the fitted parameters and their associated confidence intervals according to the following formula [15]: vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi uX u q ! 2 2 t Dt ¼ f p Dpj ; qpj j
Male Wistar rats (n = 7) underwent permanent middle cerebral artery occlusion (MCAO) to induce focal stroke [5]. Animals were euthanized by intravenous injection of saturated KCl at the end of the experiment. Animal procedures were conducted according to European Community Council Directives 86/609/EEC guidelines and approved by the Animal Care and Use Committee of the University of Eastern Finland. Magnetic resonance imaging
MRI data were acquired according to methods described by Jokivarsi et al. [5]. The trace of diffusion tensor (DAV= 1/3 trace [D]) images, Τ1ρ and T2 were obtained at 30–60 min intervals at 4.7 T for 6 h after induction of focal brain ischaemia. Five echo times (10, 20, 40, 60 and 80 ms) and five spin-lock pulse durations (10, 20, 40, 60 and 80 ms) were used for T2 and Τ1ρ MRI acquisitions, respectively. Total acquisition time for DAV, T2 and Τ1ρ was 24 min. Image computation
Image postprocessing was performed using the MRI data analysis software Mango (Research Imaging Institute, UT Health Science Center at San Antonio, Texas, USA) and software written in-house using the Matlab programming environment (MathWorks, Natick, Massachusetts, USA). Images used for the T1ρ and T2 signal intensity analyses were the sum of weighted images acquired at each echo time (T2) and spin-lock duration (T1ρ). T1ρ and T2 relaxation maps were computed using a monoexponential fit in a logarithmic space. To eliminate inter-subject variation, changes in relaxation times and absolute signal intensities were quantified relative to the nonischaemic hemisphere. Thereby each rat served as its own control. To compare ischaemic and nonischaemic hemispheres absolute percentage difference (APD) images were computed for weighted images and relaxometry maps for each rat and time point. Where entire images were mirrored about the vertical axis and manually coregistered to their originals, followed by the calculation: APD = 100 × [(original − mirrored)/(original + mirrored)].
! where f p is the model as function of fitted parameters ! p and Δpj is the confidence interval for parameter j. DAV defined ROIs were loaded onto confidence interval maps. Pixels within the ROI were colour-coded according to their confidence interval. The per cent of pixels with confidence intervals more than 95% within the ROI were calculated. To determine the relationship between stroke duration and the per cent of significant pixels within the ROI, Pearson’s correlations were conducted. The above calculation for estimating uncertainty associated with estimates of stroke duration was used.
Results ΔT1ρ or ΔT2 relaxation times and signal intensities show a significant positive correlation with stroke duration allowing stroke onset to be estimated (Fig. 1). Relaxation times provide a substantially more accurate estimate of stroke duration
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
1182 NeuroReport 2014, Vol 25 No 15
compared to signal intensities. For example, using relaxation times an estimate of 4.5 h has an uncertainty of ± 12 min for ΔT1ρ, ± 35 min for ΔT2, compared with ± 59 and ± 53 min using respective ΔT1ρ and ΔT2 signal intensities.
The amount of pixels with significant changes in T1ρ and T2 within DAV defined ROIs increases with time (Fig. 2). Both increases and decreases are seen in T2 relaxation times (Fig. 2c), but not in T1ρ. It is also evident that
Fig. 1
(a)
(b) 500
400
400
Time post-MCAO (min)
300 300
200
200
100 100 0
0
100 0
2
6 4 8 ΔT1ρ absolute relaxation times
10
12
Time post-MCAO (min)
(c) 400
(d) 400
300
300
200
200
100
100
0
−4
−2
0 6 0 2 4 ΔT2 absolute relaxation times
8
Individual rats
10
−2
−4
0
−2
1σ error bars
2 4 6 ΔT1ρ signal intensities
0
2 4 6 ΔT2 signal intensities
8
8
10
10
Linear trend line
ΔT1ρ and ΔT2 relaxation times and signal intensities as a function of ischaemia duration and linear trend lines which estimated time since middle cerebral artery occlusion (MCAO). Data are averaged over the regions of interest for each rat at each time point and the fitting performed for the pooled data over all rats. (a) ΔT1ρ relaxation times, MCAO = 29.9 (± 1.4) × ΔT1ρ, r = 0.96, P < 0.001. (b) ΔT1ρ signal intensities, MCAO = 30.3 (± 7.3) × ΔT1ρ + 69.4 (± 33.4), r = 0.81, P < 0.001. (c) ΔT2 relaxation times, MCAO = 24.1 (± 3.7) × ΔT2 + 86.7 (± 20.5), r = 0.90, P < 0.001. (d) ΔT2 signal intensities, MCAO = 22.2 (± 6.5) × ΔT2 + 117.3 (± 29.0), r = 0.75, P < 0.001.
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Timing the ischaemic stroke by 1H-MRI Rogers et al. 1183
Fig. 2
(a) Time post-MCAO ± 4 min
DAV
(b)
(c)
T1 Signal intensities
Relaxometry maps
T2 Signal intensities
Relaxometry maps
1h
2h
3h
4h
5h
6h
Confidence intervals: 99 - 100% 95 - 90% 90 - 95%
Negative difference Positive difference
≤ 90% DAV and confidence interval maps of one representative rat during the progression of ischaemia. Confidence interval maps are overlaid with weighted images and relaxometry maps for clarity. (a) DAV images used to define the region of interest (ROI). (b) Significant ΔT1ρ within DAV defined ROIs for signal intensities and relaxation times. (c) Significant ΔT2 within the DAV defined ROIs for signal intensities and relaxation times, the latter showing positive and negative ΔT2.
signal intensities demonstrate an ambiguous effect of ischaemia over time making it unsuitable for estimating stroke onset. This was observed for all animals. For both T1ρ and T2 the per cent of significant pixels (P < 0.05) within DAV defined ROIs shows a significant positive correlation with time post-MCAO allowing onset time to be estimated (Fig. 3). For example, an estimate of 4.5 h has an uncertainty of ± 19 min for T1ρ and ± 26 min for T2. To assess image quality, signal-to-noise ratio SNR : mcontralateral signal= mcontralateral noise and contrast-to mischaemic signal mcontralateral signal p ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi noise ratio CNR : were calcu2 2 nischaemic noise þncontralateral noise
lated across all rats at 4.5 h and local noise was estimated using a 5 × 5 kernel. SNR for relaxometry maps (T1ρ: 19.1, SD = 2.2; T2: 17.5, SD = 2.2) and summedweighted images (T1ρ: 18.3, SD = 1.2; T2: 17.5, SD = 1.2) were comparable, demonstrating similar SNR
for weighted images and relaxometry maps. CNR values for relaxometry maps (T1ρ: 2.8, SD = 0.5; T2: 2.0, SD = 0.7) were greater than for weighted images (T1ρ: 0.7, SD = 0.9; T2: 0.4, SD = 1.1). Average relaxation times from an ROI placed in the striatum of the nonischaemic hemisphere of scans acquired at the first time point were: T2 = 56.5 ms (SD = 1.7) and T1ρ = 74.3 ms (SD = 2.3).
Discussion The present study shows that quantification of the absolute difference in relaxation times between the ischaemic and contralateral hemisphere provides a more accurate estimate of ischaemia duration compared with respective signal intensity differences. We show that the margin of error associated with relaxation time estimates is far smaller than when using signal intensities. Issues using signal intensities for assessment of brain tissue status have been documented [13,16]. Kavec et al. [16] reported that following 30 min MCAO in rats,
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
1184
NeuroReport 2014, Vol 25 No 15
Fig. 3
(b)
(a) 400
400
Time post-MCAO (min)
300 300
200 200
100
100
0
0 0
20
40
60
80
Per cent of DAV ROI with significant ΔT1
100
0
20
40
60
80
100
Per cent of DAV ROI with significant ΔT2
The per cent of pixels within the DAV defined region of interest (ROI) with significant ΔT1ρ and ΔT2 relaxation times as a function of ischaemia duration. (a) MCAO = 3.6(± 0.3) × ΔT1ρ, r = 0.89, P < 0.001. (b) MCAO = 4.6(± 0.4) × ΔT2, r = 0.79, P < 0.001. MCAO, middle cerebral artery occlusion.
ischaemic damage was not evident at 24 h in weighted images but could be identified with absolute T1ρ and T2. Furthermore, quantification of T2-FLAIR signal intensities were found to have low sensitivity leading patients within the treatment-window to be misclassified [13]. In addition to higher CNR, relaxation times may provide a more accurate estimate of stroke onset as inherent variation due to B0 and B1 inhomogeneities are eliminated by fitting signal intensities to the monoexponential relaxation equation. One could potentially improve contrast due to ischaemia in weighted images by optimizing either the echo or spin-lock time; however, to offset the worsened SNR in required long TE or spin-lock time images, scan times may increase. A shortening of T2 was observed during the initial stages of ischaemia (Fig. 2c), an effect presumably dominated by the negative blood–oxygenation level-dependent (BOLD) effect [3,17]. At a certain time point T2 therefore, displays no difference to the contralateral hemisphere and could confound estimates of ischaemia duration. This however, occurs in the first 2 h of the insult when ischaemia may be reversible. Thomalla et al. [11] found patients without T2-FLAIR hyperintensity in DWI defined ischaemic tissue had stroke duration of less than 3 h. Thus, patients showing no significant difference in T2 relaxation times due to T2 cross-over may be within the recanalization treatment-window. T1ρ does not
shorten during ischaemia due to low sensitivity of the BOLD effect [18] and present results show it provides a more accurate estimate of stroke duration than ΔT2. Nevertheless, T1ρ is not widely used in clinical neuroMRI due to associated SAR issues. Recent advancements in MRI pulse sequencing suggest that through use of adiabatic T1ρ, imaging of the brain with good volume coverage is feasible with greatly reduced SAR both for preclinical [19] and clinical applications [20]. It has been suggested that patients with hypoperfused tissue may still benefit from thrombolysis [21]. Severe ADC decreases are typically assumed to represent irreversible tissue damage, for instance, 20% decrease in ADC of 7 h of permanent MCAO has been observed to result in tissue damage as revealed by histology [7]. However, clinical data suggest that severely reduced ADC in acute phase overestimates infarct volume [22]. Hohn-Berlage et al. [7] reported that in the early moments of ischaemia the area of elevated T2 relaxation time is smaller than the area of reduced ADC. We extended this observation using statistical levels of significance and found that the DAV defined ROIs contain pixels with normal appearing T1ρ and T2 relaxation. These may or may not represent potentially reversible ischaemic tissue. The discrepancy between DAV and T1ρ and T2 potentially indicates transition to irreversible ischaemia takes place with time. In addition, the linear
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Timing the ischaemic stroke by 1H-MRI Rogers et al. 1185
relationship between ischaemia duration and the per cent of pixels with significant ΔΤ1ρ and ΔΤ2 in the DAV ROI provides a further estimate of onset time. In this instance it should be borne in mind that each MRI pixel contains a number of neural cell types that vary regionally and have inherently different sensitivity to ischaemia. Indeed, recent data from rodent stroke models point to variation in MR relaxation time responses both in size and kinetics during ischaemia within low ADC volume [7,19].
8
9
10
11
Conclusion The present data demonstrate that absolute relaxation times are more accurate than signal intensities for MRI assessment of ischaemic stroke, both in estimating onset time and identifying ischaemic tissue. Given this, and that acquisition time would be reasonable in clinical settings (< 15 min), we encourage the use of absolute relaxation times in assessment of acute ischaemic stroke.
12
13
14
Acknowledgements This study was supported by grants from the Academy of Finland and the Sigrid Juselius Foundation.
15 16
Conflicts of interest
There are no conflicts of interest.
References 1 Hacke W, Kaste M, Bluhmki E, Brozman M, Dávalos A, Guidetti D, et al. Thrombolysis with alteplase 3 to 4.5 h after acute ischemic stroke. N Engl J Med 2008; 359:1317–1329. 2 Jones SC, Kharlamov A, Yanovski B, Kim DK, Easley KA, Yushmanov VE, et al. Stroke onset time using sodium MRI in rat focal cerebral ischemia. Stroke 2006; 37:883–888. 3 Gröhn OHJ, Kettunen MI, Mäkelä HI, Penttonen M, Pitkänen A, Lukkarinen JA, Kauppinen RA. Early detection of irreversible cerebral ischemia in the rat using dispersion of the magnetic resonance imaging relaxation time, T1p. J Cereb Blood Flow Metab 2000; 20:1457–1466. 4 Knight RA, Dereski MO, Helpern JA, Ordidge RJ, Chopp M. Magnetic resonance imaging assessment of evolving focal cerebral ischemia. Comparison with histopathology in rats. Stroke 1994; 25:1252–1261. 5 Jokivarsi KT, Hiltunen Y, Gröhn H, Tuunanen P, Gröhn OHJ, Kauppinen RA. Estimation of the onset time of cerebral ischemia using T1p and T2 MRI in rats. Stroke 2010; 41:2335–2340. 6 Siemonsen S, Mouridsen K, Holst B, Ries T, Finsterbusch J, Thomalla G, et al. Quantitative T2 values predict time from symptom onset in acute stroke patients. Stroke 2009; 40:1612–1616. 7 Hoehn-Berlage M, Eis M, Back T, Kohno K, Yamashita K. Changes of relaxation times (T1, T2) and apparent diffusion coefficient after permanent middle cerebral artery occlusion in the rat: temporal evolution, regional extent, and comparison with histology. Magn Reson Med 1995; 34:824–834.
17
18
19
20
21
22
Hoehn-Berlage M, Norris D, Kohno K, Mies G, Leibfritz D, Hossmann K. Evolution of regional changes in apparent diffusion coefficient during focal ischemia of rat brain: the relationship of quantitative diffusion NMR imaging to reduction in cerebral blood flow and metabolic disturbances. J Cereb Blood Flow Metab 1995; 15:1002–1011. Hasegawa Y, Fisher M, Latour LL, Dardzinski BJ, Sotak CH. MRI diffusion mapping of reversible and irreversible ischemic injury in focal brain ischemia. Neurology 1994; 44:1484–14890. Madai V, Galinovic I, Grittner U, Zaro-Weber O, Schneider A, Martin S, et al. DWI intensity values predict FLAIR lesions in acute ischaemic stroke. PLoS One 2014; 9:e92295. Thomalla G, Cheng B, Ebinger M, Hao Q, Tourdias T, Wu O, et al. DWIFLAIR mismatch for the identification of patients with acute ischaemic stroke within 4·5 h of symptom onset (PRE-FLAIR): a multicentre observational study. Lancet Neurol 2011; 10:978–986. Ebinger M, Galinovic I, Rozanski M, Brunecker P, Endres M, Fiebach JB. Fluid-attenuated inversion recovery evolution within 12 h from stroke onset: a reliable tissue clock? Stroke 2010; 41:250–255. Cheng B, Brinkmann M, Forkert ND, Treszl A, Ebinger M, Köhrmann M, et al. STIR and VISTA Imaging Investigators. Quantitative measurements of relative fluid-attenuated inversion recovery (FLAIR) signal intensities in acute stroke for the prediction of time from symptom onset. J Cereb Blood Flow Metab 2013; 33:76–84. Xu X, Zu Q, Lu S, Cheng Q, Yu J, Sheng Y, et al. Use of FLAIR imaging to identify onset time of cerebral ischaemia in a canine model. AJNR Am J Neuroradiol 2014; 35:311–316. Ku H. Notes on the use of propagation of error formulas. J Res Natl Bur Stand (1934) 1966; 70:263–273. Kavec M, Gröhn OHJ, Kettunen MI, Silvennoinen MJ, Garwood M, Kauppinen RA. Acute cerebral ischemia in rats studied by Carr-Purcell spinecho magnetic resonance imaging: assessment of blood oxygenation leveldependent and tissue effects on the transverse relaxation. Magn Reson Med 2004; 51:1138–1146. Calamante F, Lythgoe MF, Pell GS, Thomas DL, King MD, Busza AL, et al. Early changes in water diffusion, perfusion, T1, and T2 during focal cerebral ischemia in the rat studied at 8.5 T. Magn Reson Med 1999; 4:479–485. Kettunen MI, Gröhn OH, Penttonen M, Kauppinen RA. Cerebral T1rho relaxation time increases immediately upon global ischemia in the rat independently of blood glucose and anoxic depolarization. Magn Reson Med 2001; 46:565–572. Jokivarsi KT, Niskanen J-P, Michaeli S, Gröhn HI, Garwood M, Kauppinen RA, Gröhn OH. Quantitative assessment of water pools by T1ρ and T2 MRI in acute cerebral ischemia of the rat. J Cereb Blood Flow Metab 2009; 29:206–216. Andronesi OC, Bhat H, Reuter M, Mukherjee S, Caravan P, Rosen BR. Whole brain mapping of water pools and molecular dynamics with rotating frame MR relaxation using gradient modulated low-power adiabatic pulses. Neuroimage 2014; 89:92–109. Albers GW, Thijs VN, Wechsler L, Kemp S, Schlaug G, Skalabrin E, et al. DEFUSE Investigators. Magnetic resonance imaging profiles predict clinical response to early reperfusion: the diffusion and perfusion imaging evaluation for understanding stroke evolution (DEFUSE) study. Ann Neurol 2006; 60:508–517. Fiehler J, Foth M, Kucinski T, Knab R, von Bezold M, Weiller C, et al. Severe ADC decreases do not predict irreversible tissue damage in humans. Stroke 2002; 33:79–86.
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
