Brain Struct Funct (2012) 217:667–675 DOI 10.1007/s00429-011-0363-4
TECHNICAL NOTE
Evaluation of methods for detecting perfusion abnormalities after stroke in dysfunctional brain regions Regine Zopf • Uwe Klose • Hans-Otto Karnath
Received: 23 August 2011 / Accepted: 10 November 2011 / Published online: 29 November 2011 Ó Springer-Verlag 2011
Abstract Commonly, in lesion-behaviour studies structural changes in brain matter are depicted and analysed. However, in addition to these structural changes, brain areas might be structurally intact but non-functional due to malperfusion. These changes may be detected using perfusion-weighted MRI (PWI). Perfusion parameters most commonly used [e.g. time-to-peak (TTP)] are semi-quantitative and perfusion is evaluated in relation to a nonaffected reference area. Traditionally, the mean of a larger region in the non-affected hemisphere or the cerebellum has been used [‘‘mean contra-region of interest (ROI) comparison’’]. Our results suggest that this method is prone to biases (in particular in periventricular regions) because perfusion differs between different parts of the brain, for example, between grey and white matter. We reduced such potential biases with voxelwise inter-hemispheric comparisons. Each voxel is compared with its homologous
voxel and thus white matter with white matter and grey matter with grey matter. This automated method seems to correspond with results deriving from manual delineation of perfusion deficits. The TTP delay maps with a threshold of 3 s seem to be best comparable to manual delineation. Our method avoids the observer-dependent choice of a reference region and involves the spatial normalisation of perfusion maps. It is well suited for whole-brain analysis of abnormal perfusion in neuroscience studies as well as in clinical contexts. Keywords Perfusion-weighted imaging Magnetic resonance imaging Lesion-behaviour analysis Malperfusion Stroke
Introduction
R. Zopf (&) ARC Centre of Excellence in Cognition and its Disorders, Macquarie Centre for Cognitive Science, Macquarie University, Sydney, NSW 2109, Australia e-mail:
[email protected] U. Klose Department of Diagnostic and Interventional Neuroradiology, University of Tu¨bingen, Tu¨bingen, Germany H.-O. Karnath Center of Neurology, Division of Neuropsychology, Hertie-Institute for Clinical Brain Research, University of Tu¨bingen, Tu¨bingen, Germany H.-O. Karnath Department of Psychology, University of South Carolina, Columbia, SC, USA
The method of lesion mapping has been widely applied to uncover anatomo-behavioural relationships in the human brain (see Rorden and Karnath 2004 for a review). Structural lesions are analysed in order to identify the non-functional brain areas in patients with a particular behavioural deficit. However, structural lesion scans may not depict all non-functional tissue. It is possible that beyond those areas with a structural defect, additional (structurally intact) brain areas are non-functional because they do not receive adequate blood supply. These areas represent zones that are receiving enough oxygen to remain structurally intact, but not enough to function normally (Schlaug et al. 1999). Such areas might also contribute to the behavioural deficit and thus need to be detected, aiming to describe precise anatomo-behavioural relationships in the human brain.
123
668
Dynamic susceptibility contrast perfusion-weighted magnetic resonance imaging (DSC-PWI) can be employed to depict malperfused brain areas. This technique involves the application of a contrast material and repetitive fast brain volume acquisition. PWI complements information about irreversible damaged tissue derived from structural magnetic resonance imaging (MRI) and spiral computed tomography (spiral-CT). Lesion-behaviour studies thus increasingly consider areas with abnormal perfusion in order to study precise structure–function relationships (Hillis et al. 2001, 2002, 2005; Karnath et al. 2005; Zopf et al. 2009; Ticini et al. 2009, 2010). In order to analyse perfusion abnormalities for a patient group with a particular deficit, perfusion brain maps need to be transformed to a common standard stereotaxic space. Automated normalisation procedures are typically applied in structural lesion analyses (Rorden et al. 2011) and can also be applied for perfusion data (Karnath et al. 2005). Further, it is necessary to calculate perfusion parameters and to set parameter thresholds. One perfusion parameter which is commonly used is time-to-peak (TTP). It requires few calculations and no operator interventions, and has been shown to correlate with final stroke volume and behavioural scores (Beaulieu et al. 1999; Neumann-Haefelin et al. 1999; Hillis et al. 2001). TTP represents the time at which the largest signal drop occurs with respect to the first perfusion volume. It thus indicates the time that it takes for the contrast material and, therefore, also the blood to reach a particular part of the brain. It has been demonstrated that a delay in blood flow correlates with behavioural disorders (Neumann-Haefelin et al. 1999; Beaulieu et al. 1999). Since stenoses are known to produce falsepositive depictions of perfusion deficits, especially in TTP images (Yamada et al. 2002), in anatomo-behavioural studies it is important to investigate participants for extracranial stenoses and exclude those subjects with a relevant stenosis. Like other perfusion parameters commonly obtained with perfusion-weighted imaging, TTP is only semiquantitative or relative. Standardization across patients is, therefore, necessary to calculate delays in TTP. The most common approach for standardization of perfusion values has been to use a reference region at the same approximate position as the anticipated area of abnormal perfusion in the unaffected hemisphere [‘‘mean contra-region of interest (ROI) comparison’’]. In the case of TTP, the mean or median value calculated from this region is subtracted to obtain standardized perfusion values in the patient’s affected hemisphere. In other words, malperfusion and delays are expressed with respect to the mean value of a reference region. This method has been employed to generally depict malperfusion (Butcher et al. 2008; Dwyer et al. 2008; Neumann-Haefelin et al. 1999;
123
Brain Struct Funct (2012) 217:667–675
Schlaug et al. 1999; Singer et al. 2003) as well as to depict malperfusion in relation to cognitive function (Demeurisse et al. 1997; Hillis et al. 2001, 2002, 2005; Weiller et al. 1990, 1993). A problem of this procedure is that regions differ generally in metabolic demand. For example, Helenius et al. (2003) found that perfusion values not only differ between grey and white matter, but also between cortical and deep grey matter and between different lobes. Thus, perfusion abnormality maps based on calculations that use mean values from regions in the unaffected hemisphere necessarily lead to biases especially for large regions (Helenius et al. 2003). For example, the TTP delay values for regions with generally faster blood supply in the healthy hemisphere compared with the reference mean are underestimated, whereas delays in generally slower blood supply regions are overestimated. A further problem of standardization methods involving a reference region is that they necessitate observer-dependent decisions to define the reference region. To control for these problems, we have developed voxelwise comparisons to study the anatomy of behavioural disorders following a unilateral stroke (Karnath et al. 2005; Zopf et al. 2009; Ticini et al. 2009, 2010). This method uses voxelwise inter-hemispheric comparisons to standardize perfusion values. Each voxel is compared to its homologous voxel in the unaffected hemisphere. Therefore, standardization of perfusion values does not rely on the mean of an entire region and can be assessed without prior defining ROIs, i.e. without observer-dependent decisions. A first aim of the present study was to evaluate this new technique of voxelwise inter-hemispheric comparisons by comparing it directly with the traditional approach of mean contra-ROI comparison. Another issue of ongoing discussion related to analysis of perfusion data is what specific TTP delay is necessary to evoke a functional deficit. For example, work on crossed cerebellar diaschisis has shown that not all observed perfusion abnormalities are necessarily functionally relevant to the observed behaviour (Infeld et al. 1995). Previous research has defined the threshold for behaviourally relevant TTP delays as C3.0 s (Karnath et al. 2005; Zopf et al. 2009; Ticini et al. 2009, 2010). This was based on observations that TTP delays [2.5 s in Wernicke’s area were associated with language dysfunction (Hillis et al. 2001), and that the general functional impairment of stroke patients correlated best with the volume of perfusion abnormality for TTP delays C4.0 s (Neumann-Haefelin et al. 1999). To further investigate this issue, the second aim of the present study was to compare TTP delay maps resulting from our automated analysis method with perfusion deficit volumes obtained by visual delineation of a skilled investigator.
Brain Struct Funct (2012) 217:667–675
Methods Subjects Ten patients with acute stroke in the right hemisphere consecutively admitted to the Centre of Neurology in Tu¨bingen were included in the study. These patients had no prior history of stroke. All patients gave their informed consent to participate in the study, which was performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and was approved by the local ethics committee. Since stenoses are known to produce false-positive depictions of perfusion deficits, especially in TTP images (Yamada et al. 2002), we excluded those patients with a hemodynamically relevant stenosis in the internal carotid arteries, i.e. C70% demonstrated by Doppler sonography. Our patient sample included five male and five female patients. The age in this patient group ranged from 39 to 78 years with a mean of 59.5 years (SD = 14.0). Two of the ten participants included in the study suffered from a haemorrhage, while all others had an ischaemic stroke. MR Imaging Fifty repetitions of perfusion-weighted EPI sequences (TR 1,440 ms; TE 47 ms; FOV 230 9 230 mm2; matrix 128 9 128 pixels; 12 axial slices; slice thickness 5 mm; gap 1 mm) were obtained with 20 ml gadolinium diethyl triamineene pentaacetic acid (gd-DTPA) bolus power injected at a rate of 3–5 ml/s. The amount of bolus used depended on the body-weight of the subject. In the same scanning session also structural scans were obtained. A FLAIR sequence was acquired with 72 axial slices (thickness 1 mm; interslice gap 1 mm; a field of view (FOV) of 192 9 256 mm2, matrix 192 9 256 pixels, repetition time (TR) of 9,310 ms and an echo time (TE) of 122 ms). DWI was performed with a single-shot EPI spin echo sequence (TR 3,200 ms; TE 87 ms; FOV 230 9 230 mm2; matrix 128 9 128 pixels; slice thickness 5 mm, gap 1 mm; b values 0, 500 and 1,000 s/mm2). The mean time between stroke and imaging was 5.2 (range 0–12) days. All scans were obtained on a 1.5-T system (Magnetom Sonata, Siemens, Erlangen, Germany).
669
image and template is minimised. First, we co-registered the perfusion-weighted volumes to the DWI (b = 0) volume which is most similar to the perfusion-weighted images with respect to tissue weighting. The spatial normalisation was then carried out on the DWI image, and the obtained transformation parameters were later applied to transform the perfusion maps. For determination of the transformation parameters, cost-function masking was employed if the lesion was visible on the DWI volume (Brett et al. 2001). We used a normalisation template featuring symmetrical left–right hemispheres (AubertBroche et al. 2003). The normalised TTP maps were spatially smoothed with a Gaussian filter of 2 mm. The perfusion TTP maps were calculated using in-house software (Klose et al. 1999). Figure 1b illustrates the steps that we implemented to perform symmetric voxelwise inter-hemispheric comparisons in order to detect perfusion deficits. All image calculations were performed using the ‘imcalc’ function which is part of the SPM package. After the TTP maps were normalised, we produced mirror images which we then subtracted from the non-mirrored maps. From this corrected TTP map, we extracted perfusion maps for TTP delay thresholds of TTP C1.5 s (TTP_1.5), TTP C3 s (TTP_3), TTP C4.5 s (TTP_4.5) and TTP C6 s (TTP_6). These thresholds were defined on the basis of the perfusion scanning time resolution. Finally, we multiplied these maps with a right-hemisphere brain mask (obtained for each patient from the DWI image, b = 1,000) in order to exclude voxels that did not lie within the right hemisphere (it excluded also voxels within ventricles). To compare the automated method for perfusion analysis (Fig. 1) with the result of visual inspection by a skilled investigator, perfusion abnormalities were manually delineated on each non-normalised TTP map for each patient using the MRIcron software (Rorden et al. 2007; http://www.mricro.com). These volumes of interest (VOIs) were then normalised together with the corresponding maps. For mean contra-ROI comparisons, we used the normalised volume of interest which was manually delineated; we mirrored it and obtained the mean TTP value for this area in the left hemisphere. The mean was subtracted from the right hemisphere, i.e. the hemisphere affected by stroke. We then extracted perfusion TTP delay maps and applied the same right-hemisphere brain mask used for voxelwise comparisons.
MR analysis Figure 1a illustrates the steps that we used to obtain spatially normalised perfusion maps. The fifty perfusionweighted volumes were first spatially realigned and then transferred into stereotaxic space using the spatial normalisation algorithm provided by SPM (http://fil.ion.ucl. ac.uk/spm/) in which the sum of square difference between
Results In Fig. 2, the group overlay maps of perfusion maps obtained by manual delineation are illustrated together with the maps obtained from automated voxelwise inter-hemispheric comparisons. The TTP delay threshold C3 s seems
123
670
Brain Struct Funct (2012) 217:667–675
Fig. 1 Representation of steps for image processing to obtain normalised perfusion maps (a) and perfusion volume extraction (b) using symmetric voxelwise comparisons
to best correspond to perfusion deficits delineated by hand, whereas the TTP_1.5 maps appear generally to be larger and the TTP_4.5 and TTP_6 maps to be reduced as compared to manual delineation. For each TTP delay threshold, we also determined the size of obtained malperfusion volume for each participant. The median values are depicted in Fig. 3. Statistical analysis confirmed that the volume size differed between TTP delay thresholds (Friedman v2 = 30, df = 3, p B 0.00001). We used Wilcoxon signed-rank tests to compare the median for each TTP delay threshold to the volume size obtained through visual delineation. The TTP_1.5 maps were significantly larger than what was obtained by visual delineation (z = -2.80, p = 0.005). In contrast, maps with TTP_6 maps were significantly smaller (z = -2.07, p = 0.038) and a
123
trend for smaller TTP_4.5 maps (z = -1.68, p = 0.093) was found. Importantly, the TTP_3 maps did not differ significantly from visual-delineation maps (z = -0.153, p = 0.88). This statistical result corresponds with the visual inspection of the overlay maps (cf. Fig. 2). To compare the new technique of voxelwise interhemispheric comparisons with the traditional approach of ‘‘mean contra-ROI comparison’’, we depicted the overlay maps as well as subtraction maps of abnormal perfusion obtained from both methods for TTP_3 (Fig. 4) (please note that we found the same pattern of results across all TTP delay maps). Table 1 presents the median malperfusion volume size for this method and voxelwise interhemispheric comparisons. The voxelwise comparison volumes were statistically not different from the volumes
Brain Struct Funct (2012) 217:667–675
671
Fig. 2 Overlay maps showing the common regions of PWI abnormalities for different TTP delay threshold values in comparison to visual delineation for the group of ten patients. The number of overlapping areas with abnormal perfusion is illustrated in different colours, coding increasing frequencies with violet (n = 1) to red (n = 10). MNI z-coordinates of the transverse sections are given. The right hemisphere is presented on the right
Fig. 3 Median size for malperfusion volumes for different TTP delay thresholds and volumes obtained through manual visual delineation. Error bars indicate the interquartile range (IQR)
obtained using mean contra-ROI comparison; there was, however, a trend towards larger malperfusion volumes for mean contra-ROI comparison (Wilcoxon signed-rank tests: z = 1.68, p = 0.093). Further, we calculated the percentage of abnormal perfusion volume of each method which was also depicted in the visual-delineation map (‘‘visual-delineation congruency’’). A value of 100% indicates that all obtained abnormal perfusion voxels were also depicted in visualdelineated abnormal perfusion maps, whereas a lower value means that additional areas were depicted in the obtained malperfusion volume. As seen in Table 1, visualdelineation congruency was significantly higher for the
123
672
Brain Struct Funct (2012) 217:667–675
Fig. 4 a Overlay maps showing the common regions of perfusion abnormalities for different comparison methods for the group of ten patients. The number of overlapping areas with abnormal perfusion is illustrated in different colours, coding increasing frequencies with violet (n = 1) to red (n = 10). b Subtraction overlay maps showing the relative depiction of perfusion abnormalities (bins of 20%) of each voxel in the mean contraROI method compared to voxelwise comparisons. The colour scale covers a range of relative differences of abnormal perfusion depiction for the two methods, from more frequent depiction of abnormal perfusion for the mean contra-ROI method (positive values) to more frequent abnormal perfusion depiction in the voxelwise comparisons method (negative values). Regions with small relative differences between -20 and 20% are not depicted in this figure. MNI z-coordinates of the transverse sections are given. The right hemisphere is represented on the right
Table 1 Contrast different comparison methods Voxelwise comparisons
Mean contra-ROI comparison
Malperfusion volume
86.85 ccm
96.25 ccm
Median (interquartile range)
(54.50)
(49.95)
Visual-delineation congruency
70.40%
64.31%
Median (interquartile range)
(24.86)
(26.50)
voxelwise comparison method as compared to the mean contra-ROI method (z = -2.803, p = 0.005). Figure 4b illustrates the differences of the overlap images between the two methods. The red colours depict voxels which are relatively more often marked as ‘‘abnormal perfusion’’ using the mean contra-ROI method as compared to voxelwise comparisons. These voxels seem to concentrate in periventricular areas. Some voxels were also relatively more depicted as abnormally perfused using voxelwise comparisons as compared to the mean contraROI method (blue colours). However, these seemed to be
123
more scattered. By comparison of Figs. 2 and 4, it can be seen that periventricular areas were also less identified as ‘‘abnormally perfused’’ by visual delineation as compared to using the mean contra-ROI method.
Discussion The overlay of individual structural lesion images as well as images of functional defects due to malperfusion from a patient group with a particular behavioural deficit allows precise identification of the region(s) that are required for a particular brain function. In order to analyse perfusion abnormalities for a patient group, perfusion brain maps have to be spatially comparable and thus need to be transformed to a common standard stereotaxic space. Automated normalisation procedures are typically applied in structural lesion analyses (Rorden et al. 2011). We used similar procedures for perfusion images and voxelwise inter-hemispheric comparisons for depiction of abnormal perfusion.
Brain Struct Funct (2012) 217:667–675
The comparison of the method of voxelwise interhemispheric comparisons with the ‘‘mean contra-ROI comparison’’ method revealed that the latter approach tends to lead to larger depictions of abnormal perfusion. In particular, in contrast with results obtained from using visual delineation, these volumes seem to include more non-random noise, especially in periventricular areas. This can also be observed in previous studies which employed this method (e.g. Hillis et al. 2002; Neumann-Haefelin et al. 1999). We think that perfusion of periventricular white matter is likely the cause for these systematic biases. White matter around the ventricles is depicted as relatively malperfused as compared to the mean from a contralateral reference region (ROI method) but not compared to white matter around ventricles in the contralateral hemisphere (voxelwise method). The reason for this might be that TTP values in periventricular areas are typically increased (see Fig. 5 for an example). White matter is generally less perfused as compared to grey matter (Helenius et al. 2003). Since the reference regions involve grey matter, a systematic bias may be obtained using the mean contra-ROI method when analysing abnormal perfusion for the whole brain. In contrast, voxelwise inter-hemispheric comparisons avoid this potential problem. Since each voxel is compared with the homologous voxel in the hemisphere not affected by stroke, grey matter is thus compared to grey matter and white matter to white matter, etc. We would like to stress that the finding that our method of voxelwise comparisons is more similar to visual delineation as compared to the contralateral ROI method does not in itself suggest that the voxelwise comparisons method is more superior. The main reason is that visual delineation
Fig. 5 Example of TTP Map. Increased TTP values (lighter grey shades) can be observed around ventricles on both hemispheres as compared to the rest of the brain. The right hemisphere is presented on the right
673
cannot be used as a ‘‘gold standard’’. Visual delineation of perfusion deficits relies on observer-dependent decisions, it has been related to poor interrater reliability (Coutts et al. 2003), and it does not make use of specific perfusion thresholds (e.g. TTP delays C3 s). However, the significant difference we found (Figs. 3, 4) points to a difference between the voxelwise and the ROI comparison methods, and suggests that the voxelwise comparison method may reduce the amount of systematic artefacts. Further, the present results confirmed that our automated method (Fig. 1) can be used successfully to extract perfusion deficits and that the TTP_3 maps seems to be best comparable to visually estimated perfusion deficits. This is an interesting new observation. Previous studies suggested the use of such threshold (TTP delay threshold C3 s) to differentiate between functional and non-functional tissue based on the finding that TTP delays [2.5 s in Wernicke’s area were associated with language dysfunction (Hillis et al. 2001), and that the general functional impairment of stroke patients correlated best with the volume of perfusion abnormality for TTP delays C4.0 s (Neumann-Haefelin et al. 1999). Again we would like to stress that visual delineation of perfusion deficits should not be seen as a ‘‘gold standard’’. However, since many previous studies have employed visual delineation (e.g. Beaulieu et al. 1999), the present results suggest that the used volumes in those previous studies represent TTP values with approximate delay thresholds of C3 s. Until more is known about TTP delays in relation to cognitive function and dysfunction, we thus recommend using a TTP threshold value of C3 s as a working model. However, more studies are needed to determine behavioural relevant perfusion thresholds. These may differ between brain areas and possibly also between observable symptoms. Furthermore, differences in the effects of malperfusion may also exist between different causes for stroke, for example, ischaemia and haemorrhage. In this study, we did not explicitly differentiate different causes, including participant with malperfusion of structurally intact tissue due to both ischaemia (n = 8) as well as haemorrhage (n = 2). One general disadvantage of perfusion thresholding approaches (both voxelwise and mean ROI approaches) is that they commonly produce some grainy looking noise, i.e. they depict voxels as malperfused which are in fact not. This is due to the low quality of perfusion scans. We slightly smoothed perfusion maps to reduce this noise, but there was still a considerable amount of random noise present in the overlay maps (see Fig. 2). However, when overlaying several perfusion maps in group studies this noise should cancel out. Furthermore, it may be possible to add automatic algorithms which consider, for example, only contiguous tissue (Dwyer et al. 2008).
123
674
Our method is excellently suited for combined analyses of abnormally perfused brain areas and structural lesions. Researchers and clinicians can perform simple image calculations with malperfusion volumes and structural damage volumes. This allows visualization of only those areas in the brain that are for example structurally intact but malperfused (Karnath et al. 2005; Ticini et al. 2009, 2010; Zopf et al. 2009). In summary, we present a method which can be used to analyse perfusion data in group lesion studies to investigate which brain areas need to be functional for a certain behavioural task. This method involves the normalisation of perfusion maps and voxelwise inter-hemispheric comparisons to obtain maps which depict perfusion deficits. Voxelwise inter-hemispheric comparisons have important advantages to calculating regional means, as regional differences and potential difference between grey and white matter are better accounted for. Our method avoids averaging procedures to obtain a reference in the healthy hemisphere, and thus also the problem of a priori anatomical region delineation and possible observer-dependent biases. In principle, this method can be applied for perfusion parameters other than TTP. Our method is not only interesting for research in cognitive neuroscience, but likewise may also be applied for assessment and quantification of perfusion deficits in a clinical context. Acknowledgments This work was supported by the Deutsche Forschungsgemeinschaft (KA 1258/10-1). We thank Leif Johannsen, Monika Fruhmann Berger and the team of the Division of Neuropsychology for their help in data acquisition and clinical assessments as well as for helpful discussions.
References Aubert-Broche B, Grova C, Jannin P, Buvat I, Benali H, Gibaud B (2003) Detection of inter-hemispheric asymmetries of brain perfusion in SPECT. Phys Med Biol 48(11):1505–1517 Beaulieu C, de Crespigny A, Tong DC, Moseley ME, Albers GW, Marks MP (1999) Longitudinal magnetic resonance imaging study of perfusion and diffusion in stroke: evolution of lesion volume and correlation with clinical outcome. Ann Neurol 46:568–578 Brett M, Leff AP, Rorden C, Ashburner J (2001) Spatial normalization of brain images with focal lesions using cost function masking. Neuroimage 14(2):486–500 Butcher K, Parsons M, Allport L, Lee SB, Barber PA, Tress B, Donnan GA, Davis SM (2008) Rapid assessment of perfusiondiffusion mismatch. Stroke 39(1):75–81 Coutts SB, Simon JE, Tomanek AI, Barber PA, Chan J, Hudon ME, Mitchell JR, Frayne R, Eliasziw M, Buchan AM, Demchuk AM (2003) Reliability of assessing percentage of diffusion-perfusion mismatch. Stroke 34(7):1681–1683 Demeurisse G, Hublet C, Paternot J, Colson C, Serniclaes W (1997) Pathogenesis of subcortical visuo-spatial neglect. A HMPAO SPECT study. Neuropsychologia 35(5):731–735
123
Brain Struct Funct (2012) 217:667–675 Dwyer MG, Bergsland N, Saluste E, Sharma J, Jaisani Z, Durfee J, Abdelrahman N, Minagar A, Hoque R, Munschauer FE 3rd, Zivadinov R (2008) Application of hidden Markov random field approach for quantification of perfusion/diffusion mismatch in acute ischemic stroke. Neurol Res 30(8):827–834 Helenius J, Perkio J, Soinne L, Ostergaard L, Carano RA, Salonen O, Savolainen S, Kaste M, Aronen HJ, Tatlisumak T (2003) Cerebral hemodynamics in a healthy population measured by dynamic susceptibility contrast MR imaging. Acta Radiol 44(5):538–546 Hillis AE, Wityk RJ, Tuffiash E, Beauchamp NJ, Jacobs MA, Barker PB, Selnes OA (2001) Hypoperfusion of Wernicke’s area predicts severity of semantic deficit in acute stroke. Ann Neurol 50:561–566 Hillis AE, Wityk RJ, Barker PB, Beauchamp NJ, Gailloud P, Murphy K, Cooper O, Metter EJ (2002) Subcortical aphasia and neglect in acute stroke: the role of cortical hyperfusion. Brain 125:1094–1104 Hillis AE, Newhart M, Heidler J, Barker PB, Herskovits EH, Degaonkar M (2005) Anatomy of spatial attention: insights from perfusion imaging and hemispatial neglect in acute stroke. J Neurosci 25(12):3161–3167 Infeld B, Davis SM, Lichtenstein M, Mitchell PJ, Hopper JL (1995) Crossed cerebellar diaschisis and brain recovery after stroke. Stroke 26(1):90–95 Karnath HO, Zopf R, Johannsen L, Fruhmann Berger M, Na¨gele T, Klose U (2005) Normalized perfusion MRI to identify common areas of dysfunction: patients with basal ganglia neglect. Brain 128:2462–2469 Klose U, Na¨gele T, Friese S, Bitzer M (1999) Charakteristische Gro¨ßen bei der MR-Untersuchung der zerebralen Durchblutung mit hoher ra¨umlicher und zeitlicher Auflo¨sung. Fortschr Ro¨ntgenstr 170:474–481 Neumann-Haefelin T, Wittsack HJ, Wenserski F, Siebler M, Seitz RJ, Modder U, Freund HJ (1999) Diffusion- and perfusion-weighted MRI. The DWI/PWI mismatch region in acute stroke. Stroke 30(8):1591–1597 Rorden C, Karnath HO (2004) Using human brain lesions to infer function: a relic from a past era in the fMRI age? Nat Rev Neurosci 5(10):813–819 Rorden C, Karnath HO, Bonilha L (2007) Improving lesion-symptom mapping. J Cogn Neurosci 19(7):1081–1088 Rorden C, Bonilha L, Fridriksson J, Bender B, Karnath HO (2011) Age-adapted templates for spatial normalization of CT and MRI scans (submitted) Schlaug G, Benfield A, Baird AE, Siewert B, Lo¨vblad K-O, Parker RA, Edelman RR, Warach S (1999) The ischemic penumbra. Operationally defined by diffusion and perfusion MRI. Neurology 53:1528–1537 Singer OC, de Rochemont R, Foerch C, Stengel A, Lanfermann H, Sitzer M, Neumann-Haefelin T (2003) Relation between relative cerebral blood flow, relative cerebral blood volume, and mean transit time in patients with acute ischemic stroke determined by perfusion-weighted MRI. J Cereb Blood Flow Metab 23:605–611 Ticini LF, Klose U, Nagele T, Karnath HO (2009) Perfusion imaging in Pusher syndrome to investigate the neural substrates involved in controlling upright body position. PLoS One 4(5): e5737 Ticini LF, de Haan B, Klose U, Nagele T, Karnath HO (2010) The role of temporo-parietal cortex in subcortical visual extinction. J Cogn Neurosci 22(9):2141–2150 Weiller C, Ringelstein EB, Reiche W, Thron A, Buell U (1990) The large striatocapsular infarct. A clinical and pathophysiological entity. Arch Neurol 47(10):1085–1091
Brain Struct Funct (2012) 217:667–675 Weiller C, Willmes K, Reiche W, Thron A, Isensee C, Buell U, Ringelstein EB (1993) The case of aphasia or neglect after striatocapsular infarction. Brain 116(Pt 6):1509–1525 Yamada K, Wu O, Gonzalez RG, Bakker D, Ostergard L, Copen WA, Weisskoff RM, Rosen BR, Yagi K, Nishimura T, Sorensen AG (2002) Magnetic resonance perfusion-weighted imaging of acute
675 cerebral infarction. Effect of the calculation methods and underlying vasculopathy. Stroke 33:87–94 Zopf R, Fruhmann Berger M, Klose U, Karnath HO (2009) Perfusion imaging of the right perisylvian neural network in acute spatial neglect. Front Hum Neurosci 3:15
123
