Journal of Neuroradiology 34 (2007) 42–48
ORIGINAL ARTICLE
Hyperacute intraventricular hemorrhage: detection and characterization, a comparison between 5 MRI sequences Hémorragie intraventriculaire en phase hyperaiguë : détection et caractérisation à partir de cinq séquences IRM D. Hernalsteena, A. Dignaca, C. Oppenheimc, A. Peetersb, L. Hermoyea, T. Dupreza, G. Cosnarda,* a b c
Department of Neuroradiology, cliniques universitaires Saint-Luc, 1200 Brussels, Belgium Department of Neurology, cliniques universitaires Saint-Luc, 1200 Brussels, Belgium Department of Neuroradiology, centre hospitalier Sainte-Anne, université Paris-V, Paris, France
KEYWORDS Intraventricular hemorrhage; MR imaging; FLAIR sequence; Echo-planar imaging; Diffusion weighted imaging
Abstract We aimed to evaluate the diagnostic accuracy of MRI for detecting early intraventricular hemorrhage (IVH) (within 6 hours after hemorrhage and to describe the MR features that allow diagnosis. For this purpose, MR data of 22 patients with hyperacute intraparenchymal hemorrhage were independently rated as negative or positive for IVH by two observers, in a blind, retrospective study taking computed tomography (CT) as providing the correct diagnosis of IVH. Sensitivity, specificity, intra- and interobserver agreement were assessed. On FSEFLAIR, EPI-GRE-T2* and DWI images, all cases of IVH were correctly rated (sensitivity of 100%). For b0 EPI images, obtained from diffusion-weighted echo planar sequences, one case of IVH was missed by one reader (sensitivity of 88%). For T1 images, one patient was incorrectly rated negative for IVH by the two readers (sensitivity of 90%). Three forms of IVH were described, including clotted hematoma, layered hemorrhage and red blood cell deposit. When CT images were obtained within a time span of less than 3 hours after MRI, volume was assessed. Volume of hemorrhage on CT correlated best with DWI images but was underestimated on EPI-GRE T2* images.
© 2007 Elsevier Masson SAS. All rights reserved.
* Corresponding author. Department of Radiology and Medical Imaging, cliniques universitaires Saint-Luc, université catholique de Louvain, 10, avenue Hippocrate, 1200 Brussels, Belgium. E-mail address:
[email protected] (G. Cosnard).
0150-9861/$ - see front matter © 2007 Elsevier Masson SAS. All rights reserved. doi:10.1016/j.neurad.2007.01.001
Hyperacute intraventricular hemorrhage: detection and characterization, a comparison between 5 MRI sequences
MOTS CLÉS Hémorragie intraventriculaire ; IRM ; Séquence FLAIR ; Imagerie echoplanar ; Imagerie de diffusion
43
Résumé Le but de cette étude était d’évaluer les performances diagnostiques de l’IRM dans la détection précoce (six heures) de l’hémorragie intraventriculaire (HIV) et de décrire les critères séméiologiques d’IRM qui permettent son diagnostic. Vingt-deux patients avec hématome intracérébral ont été classés de façon indépendante et à l’aveugle comme présentant ou non une HIV par deux observateurs. Le scanner a été utilisé comme gold standard. La sensibilité, spécificité et évaluation intra- et interobservateur ont été déterminées. En séquence FLAIR, EPI-GRE-T2* et imagerie de diffusion à b1000, toutes les hémorragies intraventriculaires étaient détectées (sensibilité de 100 %). Un cas d’hémorragie intraventriculaire n’a pas été détecté par un lecteur en imagerie de diffusion à b0 (sensibilité de 88 %). En séquence T1, une hémorragie intraventriculaire n’était pas détectée par les deux lecteurs (sensibilité de 90 %). Trois formes d’hémorragie intraventriculaire étaient décrites : caillot, niveau hémorragique et sédiment globulaire. Dans les cas où un scanner était réalisé dans les trois heures suivant l’IRM, l’aspect de l’hématome était comparé. L’image de diffusion s’est avérée la plus proche de celle du scanner. En revanche, la séquence EPI-GRE-T2* sous-estimait notablement le volume de l’hématome.
© 2007 Elsevier Masson SAS. All rights reserved.
Introduction During the last decade, MR imaging has been increasingly used as the primary imaging tool for patients with suspicion of acute stroke. Although CT is still the usual means for detecting early hemorrhage, several studies support the use of early MR imaging for this purpose [1,8,12,14,16–18] owing to its high sensitivity for identifying early blood degradation products. Despite the fact that appearance and progression of intraparenchymal hemorrhage (IPH) on MR images have received extensive study over the past several years [2,3,6,7,9–11,15], the appearance of intraventricular blood, within the first few hours, has rarely come to clinical attention. In an early report, Bakshi et al. was the first group to describe two forms of intraventricular hemorrhage (IVH) by MRI, a fixed clot or a free- flowing mixture layering with cerebrospinal fluid (CSF), with or without blood -CSF level [4]. Later, they showed that FLAIR MR imaging was very sensitive to acute and subacute intraventricular hemorrhage [5]. The present study was conducted: ● to evaluate the sensitivity of MRI in the detection of hyperacute IVH, associated with deep, large hematoma, among 5 different MR sequences, using CT as the gold standard; ● and to investigate the appearance of intraventricular blood.
We tested the hypothesis that early MR imaging, using a standardized stroke protocol, would be effective and reliable in the diagnosis of hyperacute IVH with high sensitivity and specificity.
Material and methods Study population After a retrospective review of the stroke unit reports of our hospital between November 1997 and April 2004, two-
hundred and five consecutive patients (N = 205) with acute stroke were identified. Only those with IPH who were referred for MR examination in the setting of acute stroke within 6 hours after symptoms onset and who had a CT-scan within the first 7 days confirming the diagnosis of IVH, without any clinical change or brain surgery, were included. Among the 205 patients considered, only 22 patients with IPH remained and by coincidence exactly half of them had secondary IVH. The patients without IVH were considered as control subjects.
Imaging protocols All patients underwent MRI followed by CT. The mean time between MR examination and symptom onset was 3 hours (range, 15 minutes -6 hours). The mean interval between MRI and CT was approximately 36 hours (range, 30 minutes-7 days), and 5 CT were acquired less than 3 hours after MRI. MRI imaging was performed on a 1.5T system (GE unit, Milwaukee, WI) with echo planar imaging capability (EPI), using 5-mm sagittal T1- (TR/TE 300/2,9ms; matrix 256 × 160; imaging time, 50 s), 5-mm axial fast spinecho fluid-attenuated inversion recovery (FSE- FLAIR)(TR/ TE/TI 10002/140/2200 ms; matrix 256 × 160; imaging time 3 min 1 s), echo- planar imaging gradient-recalled echo (EPI GRE T2*)- (TR/TE 4200/50 ms; matrix 128 × 128; imaging time: 59 s) and axial diffusion weighted imaging (DWI) using spin-echo EPI, and two b values, including b = 0 s/mm2 and b = 1000s/mm2 (TR//TE 4500/94 ms; matrix 96 × 64; imaging time: 36 s). Most of the time, this whole standardized MR protocol for hyperacute stroke was not performed because the diagnosis of IPH was obvious. Overall, 20 patients had sagittal GRE-T1, 11 axial FLAIR, 21 GRE EPI-T2* and 13 DWI sequences were obtained. Of the 11 patients with IVH, 10 T1, 3 FLAIR, 11 GRE EPI-T2* and 8 DWI sequences were obtained. To qualify for enrollment, at least two sequences had to be completed. Computed tomographic scans were performed on 1 of the following scanners: MX 8000 (Marconi), DUAL and Bril-
44
D. Hernalsteen et al.
lance (Philips). Images were acquired in the orbito-meatal plane with 5 mm thickness.
Data collection and Processing The studies were reviewed retrospectively by two neuroradiologists (D.H., reader 1, T.D. reader 2) blinded to the clinical information. All MR sequences from each subject were randomly ordered and interpretations were performed on different days to avoid reader recognition. They were independently rated on a 6-level scale: [1] no IVH (with high certitude), [2] no IVH (with medium certitude), [3] no IVH (with low certitude), [4] IVH (with low certitude), [4] IVH (with medium certitude), [6] IVH (with high certitude). For sensibility and specificity analysis, a binary decision was extracted from this 6-level scale: no for levels 1-4, yes for levels 5-6. The sensitivity and specificity of each MR sequence was calculated, taking the CT diagnosis as the gold standard. Inter- and intra- observer agreement were calculated for each sequence by kappa-analysis on the 6level scale. IVH was further evaluated for pattern and signal intensity for each lateral ventricle (2 descriptions/patient) by a consensus reading of two others radiologists (A.D., G.C).Third and fourth ventricles were not observed because their involvement was less common. Pattern of intraventricular blood was assessed according to the forms described by Bakshi [4], to which we added a new pattern, called red blood cell deposit (RBC deposit), corresponding to sedimentation of agglutinated red blood cells in the posterior
aspect of the ventricles. Signal intensity of IVH was compared with those of the intraparenchymal hematoma of origin and to CSF. When a CT acquired less than 3 hours after the MRI (5 patients)was available, a visual comparison was done between CT and MRI to assess the volume of IVH, and quoted as similar, underestimated or overestimated on MR.
Results Twenty-two patients with IPH, eleven of them with IVH, were retrospectively studied. There were 18 males and 4 females patients ranging in age from 51 and 89 years (mean age, 69 years). Sixteen cases involved arterial hypertension; one, arteriovenous malformation; one, low platelets; and four, unknown etiology. The accuracy of MR for identifying hyperacute IVH is shown in Tables 1,2. The results of imaging findings of IVH, analyzing pattern and signal intensitity , are summarized in Tables 3,4. FLAIR detected all cases of IVH (3 of 3), and none of the FLAIR images of the control subjects were falsely rated (100% sensitivity, 100% specificity). T1 detected 9 of the 10 cases of IVH, but for reader 2, two of the T1 images of the control subjects falsely depicted IVH (90% sensitivity, 100 (rd 1) and 80% (rd 2) specificity). EG EPIT2* detected all cases of IVH (11 of 11), but for reader 1, two of the EG EPI-T2* images of the control subjects falsely detected IVH (100% sensitivity, 80% (rd 1) and 100% (reader 2) specificity). b0 detected all cases of IVH (8 of 8) for
Table 1 Diagnostic performance figures (with 95% confidence intervals) of both readers for detection of IVH on 5 MRI sequences Tableau 1 Performance diagnostique (intervalle de confiance : 95 %) des deux lecteurs pour la détection des hémorragies intraventriculaires sur cinq séquences d’IRM Reader 1
Sensitivity Specificity
T1(N = 20) 90%(9/10) 100%(10/10)
FLAIR (N = 11) 100%(3/3) 100%(8/8)
EPI T2* (N = 21) 100%(11/11) 80%(8/10)
B0 (N = 13) 100%(8/8) 100%(5/5)
B1000 (N = 13) 100%(8/8) 80%(4/5)
Reader 2
Sensitivity Specificity
90%(9/10) 80%(8/10)
100%(3/3) 100%(8/8)
100%(11/11) 100%(10/10)
88%(7/8) 100%(5/5)
100%(8/8) 100%(5/5)
Table 2 Intra- and interobserver agreement for each MRI sequence Tableau 2 Accord intra- et interobservateur pour chaque séquence Pulse sequence T1 FLAIR EPI T2* B0 B1000
Interobserver agreement (k) 0,662 0,891 0,779 0,922 0,838
Intraobserver agreement 0,723 1 0,804 0,922 0,917
Table 3 Number and patterns of intraventricular hemorrhage identified on sequence considering two ventricles per patient (N = 11, 22 ventricles). RBC = red blood cell Tableau 3 Nombre et caractéristiques des hémorragies intraventriculaires identifiés en considérant deux ventricules par patient (n = 1 ; 22 ventricules). RBC = red blood cell Clotted Layered RBC deposit Total
T1 (N = 10) 9 5 0 14
FLAIR (N = 3) 2 3 0 5
EPI T2* (N = 11) 12 1 18 31
B0 (N = 8) 6 1 3 10
B1000 (N = 8) 8 5 0 13
Hyperacute intraventricular hemorrhage: detection and characterization, a comparison between 5 MRI sequences
45
Table 4 Signal intensity of IVH (versus intraparenchymal hematoma of origin and versus CSF) identified in each sequence Tableau 4 Intensité du signal de l’hémorragie intraventriculaire (par rapport à l’hématome intraparenchymateux d’origine et le LCS) identifiée sur chaque séquence Clotted Layered
RBC
T1 Iso/hyper vs CSF
FLAIR Iso/hyper vs CSF
Slightly hypo/slightly hyper vs CSF, with a gradient
Slightly hypo/hyper vs CSF
EPI T2* Iso/Iso CSF, with a hypo rim
B0 Iso/hypo vs CSF, with a hypo rim
B1000 Iso/hyper vs CSF Slightly hypo/hyper vs CSF
Very hypo
Very hypo
Figure 1 Hypertensive parenchymal hematoma in the left basal ganglia with intraventricular extension of clotted pattern. a. Axial CT scan obtained after MR images. b. Axial b1000 diffusion-weighted image. c. Axial GRE EPI-T2* weighted image at the same level. d. Sagittal T1 image. The clotted hematoma appears as a lesion of high signal intensity on DWI (arrowheads), of lower signal on GRE EPI-T2*, surrounded by a rim of low intensity (arrows), and of mildly high signal relative to CSF on T1. Figure 1 Hématome intracérébral basal gauche, d’origine hypertensive, avec extension intraventriculaire d’aspect cailloté. a. Scanner cérébral en coupes axiales, obtenues après l’IRM. b. Coupe IRM axiale pondérée b1000 en séquence de diffusion. c. Coupe IRM axiale pondérée EG EPI-T2* axiale au même niveau. d. Coupe IRM sagittale pondérée T1. L’hématome intraventriculaire est de signal intense en séquence de diffusion b1000 (tête de flèche), de moindre signal en séquence EPI GRE-T2*, entourée d’une coque de faible intensité (flèche), et de signal légèrement plus élevé que le LCR en séquence pondérée T1.
reade 1 but 7 of 8 for reade 2. None of the b0 images of the control subjects were falsely rated (100 and 88% sensitivity, 100% specificity). b1000 detected all cases of IVH (8 of 8)
but one of the b1000 images of the control subjects was falsely rated by rd 2 (100% sensitivity, 100 and 80% specificity). Ranges for interobserver agreement based on the ĸ
46
Figure 2 Clotted intraventricular hematoma in the right lateral ventricle. On FLAIR sequence, clotted hematoma appears as a well limited lesion of high signal intensity, similar to deep intracerebral hematoma of origin. Figure 2 Hématome intraventriculaire cailloté dans le ventricule latéral droit. En séquence FLAIR, le caillot apparaît comme une lésion bien délimitée, de signal intense, semblable à celui de l’hématome intracérébral profond d’origine.
Figure 3 Coexisting clotted and layered intraventricular hemorrhage in the right lateral ventricle. On FLAIR sequence, layered hemorrhage appears as a lesion of less high intensity than clotted hematoma and presents a fluid-level in this case. Figure 3 Hémorragie intraventriculaire dans le ventricule latéral droit, coexistence de caillot et de niveau hémorragique déclive. En séquence FLAIR, l’hémorragie déclive apparaît comme une lésion de moindre intensité que le caillot et présente un niveau liquidien déclive.
statistic for paired observers were from 0,662 to 0,922 for MRI. Intraobserver agreement varied from 0, 723 to 1. Three patterns of IVH in lateral ventricles were observed: clotted hematoma, layered hemorrhage and red blood cell deposit. These patterns co-existed frequently but had different signal characteristics. They were almost always in
D. Hernalsteen et al.
Figure 4 Red Blood Cell (RBC) Deposit. RBC deposit corresponds to little hemorrhage limited to the occipital horns of the lateral ventricles, in this case on the left side (arrow), and almost only detected by EPI GRE T2*. Figure 4 Sédiment globulaire. Le sédiment globulaire correspond à une minime hémorragie intraventriculaire, limitée aux cornes occipitales des ventricules latéraux, dans ce cas-ci à gauche (flèche), et presque exclusivement détecté en séquence EPI GRE-T2*.
the lateral ventricles, usually bilateral. Clotted hematoma, well limited, appeared as a lesion of similar appearance as parenchymal hematoma, of mildly high signal relative to CSF on T1 (Fig. 1d), of high intensity on FLAIR (Fig. 2) and DWI images (Fig. 1b, Fig. 5c). On b0 images (Fig. 5b) and EG EPI-T2*(Fig. 1c), however, it was of lower intensity than CSF, surrounded by a characteristic peripheral rim of low intensity. Layered hemorrhage appeared more as a homogenous lesion of less high intensity on T1 and FLAIR (Fig. 3) than coexisting IPH because of dilution. On EG EPI-T2* and b0 images (Fig. 5b)layered hemorrhage was difficult to detect. Red blood cell deposit appeared as a small lesion limited to the bottom of the occipital horns (Fig. 4), of deeply hypointense signal and only detected by EPI EG-T2* and sometimes by b0 images of DWI. For evaluating the volume of hemorrhage, DWI images were the most similar to CT (Fig. 1a,b and Fig. 5a,c).
Discussion In analogy to subarachnoid hemorrhage (SAH), fresh intraventricular blood degrades more slowly than coexisting intraparenchymal hematoma because dilution by CSF and higher local oxygen content [6]. Therefore, recognition of IVH has prognostic value and may help to select patients for interventional therapy to avoid obstructive hydrocephalus. There are only a few studies on MR imaging of hyperacute intraventricular hemorrhage. Bakshi was the first to analyze the MR features of IVH, to describe two major forms, clotted hematoma and layered hemorrhage [4] and later to demonstrate the high sensitivity of FLAIR for acute and subacute IVH [5]. In our study, a third form of IVH was
Hyperacute intraventricular hemorrhage: detection and characterization, a comparison between 5 MRI sequences
47
Figure 5 Clotted hematoma in the left lateral ventricle secondary to a deep hematoma, associated with a layered hemorrhage in the right ventricle. a. axial CT scan obtained after MR images; b. b0 SE EPI images; c. b1000 images. Clotted hematoma appears as a lesion of lower intensity than CSF on the b0 images, but of high intensity on the b1000 images. Layered hemorrhage is of less high intensity on b1000 and not detectable on b0 images. Figure 5 Hématome cailloté dans le ventricule latéral gauche, secondaire à un hématome intracérébral profond, associé à une hémorragie déclive dans le ventricule latéral droit. a. Coupe axiale en CT obtenu après l’IRM. b. Coupe axiale pondérée b0 en séquence de diffusion. c. Coupe axiale pondérée b1000 en séquence de diffusion. L’hématome cailloté a une intensité moindre que celle du LCR sur l’image b0, moins plus élevée à b1000. L’hémorragie déclive est de moindre intensité à b1000 et n’est pas détectable à b0.
shown, called red blood cell (RBC) deposit, corresponding to little hemorrhage limited to the bottom the ventricular horns, detected by more advanced techniques, including GRE EPI-T2* and SE EPI imaging. We observed that the 5 studied MRI sequences were able to assess the presence of intraventricular blood with high sensitivity and specificity. As expected, FLAIR and diffusion weighted imaging had the best accuracy for detecting clotted and layered hemorrhage. The usefulness of FLAIR in the diagnosis of IVH and SAH is well known [13] and relates to the hyperintense appearance of these hemorrhage, caused by a higher pro-
tein concentration, against a dark CSF background allowing blood to be seen in ventricles and sulci. On the diffusionweighted images, IVH also appears as a hyperintense lesion against a dark CSF background, but corresponds to an area of decreased diffusion. On these sequences, the volume of hemorrhage was the most similar to CT. b0 EPI and EPI-GRE T2* sequences are sensitive to the susceptibility effects of magnetic field inhomogeneity of hyperacute intracranial hemorrhage [18]. For detecting IVH however, CSF dilutes blood and therefore lowers the concentration of paramagnetic deoxyhemoglobin necessary to cause hypointensity on
48
D. Hernalsteen et al.
these sequences. For this reason, together with the high signal of surrounding CSF, clotted and layered IVH were more difficult to see. Detection of clotted hematoma was possible because of its mildly lower signal intensity than CSF and the presence of a characteristic rim of low intensity, due to susceptibility artifacts at the interface clot-CSF. However, RBC deposit was very well seen, probably explained by a higher concentration of aggregated red blood cells and the consecutive addition of paramagnetic effect of each RBC. Greater contrast resolution was achieved by GRE EPI-T2*, allowing the detection of subtle hemorrhagic lesions that may not be apparent on SE EPI T2. The weaknesses in this study include small size, failure to obtain all MR sequences and delays between CT and MR imaging. Indeed, because this study was retrospective, we did not perform a complete MR imaging protocol for every patient, which would have allowed to have a sufficient number of each sequence. Nevertheless, false negative or false positive diagnosis were certainly possible, and concerned often intraventricular hemorrhage of small size. In such cases, the diagnosis was not missed on the other sequences.
[4]
Conclusion
[11]
MRI seems to be diagnostic in the evaluation of hyperacute intraventricular hemorrhage as a complication of a deep intraparenchymal hematoma and may be of use as a primary diagnostic tool in stroke patient. FSE-FLAIR best performed for detection of clotted and layered hemorrhage and b1000 SE-DWI for quantification. EPI-GRE-T2* significantly underestimated the extent of bleeding when compared to CT. b0 images could not replace GRE images for the detection of acute blood products.
References [1]
[2]
[3]
Arnould MC, Grandin CB, Peeters A, Cosnard G, Duprez TP. Comparison of CT and three MR sequences for detecting and categorizing early (48 hours) hemorraghic transformation in hyperacute ischemic stroke. AJNR Am J Neuroradiol 2004;25: 939–44. Atlas S, Mark A, Grossman R, Gomori J. Intracranial hemorrhage: gradient-echo MR imaging at 1.5 T. Comparison with spin-echo imaging and clinical applications. Radiology 1988; 168:803–7. Atlas SW, Thulborn KR. MR detection of hyperacute parenchymal hemorrhage of the brain. AJNR Am J Neuroradiol 1998;19: 1471–7.
[5]
[6] [7]
[8]
[9]
[10]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
Bakshi R, Kamran S, Kinkel PR, Bates VE, Mechtler LL, Belani SL, et al. MRI in cerebral intraventricular hemorrhage: analysis of 50 consecutive cases. Neuroradiology 1999;41:401– 9. Bakshi R, Kamran S, Kinkel PR, Bates VE, Mechtler LL, Janardhan V, et al. Fluid-attenuated inversion-recovery MR imaging in acute and subacute cerebral intraventricular hemorrhage. AJNR Am J Neuroradiol 1999;20:629–36. Bradley WJ. MR appearance of hemorrhage in the brain. Radiology 1993;189:15–26. Clark R, Watanabe A, Bradley WJ, Roberts J. Acute hematomas: effects of deoxygenation, hematocrit, and fibrin-clot formation and retraction on T2 shortening. Radiology 1990; 175:201–6. Fiebach JB, Schellinger PD, Gass A, et al. Stroke magnetic resonance imaging is accurate in hyperacute intracerebral hemorrhage. A multicenter study on the validity of stroke imaging. Stroke 2004;22:1–5. Gomori JM, Grosman RI, Goldberg HI, Zimmerman RA, Bilaniuk LT. Intracranial hematomas: imaging by high-field MR. Radiology 1985;157:87–93. Hackney D, Atlas S, Grossman R, et al. Subacute intracranial hemorrhage: contribution of spin density to appearance on spin-echo MR images. Radiology 1987;165:199–202. Hayman L, Taber K, Ford J, Bryan R. Mechanisms of MR signal alteration by acute intracerebral blood: old concepts and new theories. AJNR Am J Neuroradiol 1991;12:899–907. Linfante I, Llinas RH, Caplan LR, Warach S. MRI features of intracerebral hemorrhage within 2 hours from symptom onset. Stroke 1999;30:2263–7. Noguchi K, Ogawa T, Inugami A, Toyoshima H, Sugawara S, Hatazawa J, et al. Acute subarachnoid hemorrhage: MR imaging with fluid-attenuated inversion recovery pulse sequences. Radiology 1995;196(3):773–4. Oppenheim C, Touzé E, Hernalsteen D, Peeters A, Lamy C, Mas J-L, Meder J-F, Cosnard G. Comparison of five MR sequences for the detection of acute intracranial hemorrhage. Cerebrovasc Dis 2005;20:388–94. Parizel PM, Makkat S, Van Miert E, et al. Intracranial hemorrhage: principles of CT and MRI interpretation. Eur Radiol 2001;11:1770–83. Patel MR, Edelman RR, Warach S. Detection of hyperacute primary intraparenchymal hemorrhage by magnetic resonance imaging. Stroke 1996;27:2321–4. Schellinger PD, Jansen O, Fiebach JB, Hacke W, Sartor K. A standardized MRI stroke protocol: comparison with CT in hyperacute intracerebral hemorrhage. Stroke 1999;30:765–8. Wiesmann M, Mayer TE, Yousry I, Hamann GF, Brückmann H. Detection of hyperacute parenchymal hemorrhage of the brain using echo-planar T2*-weighted and diffusion-weighted MRI. Eur Radiol 2001;11:849–53.
