- Email: [email protected]
Long-term visibility of primary intracerebral hemorrhage on magnetic resonance imaging
Long-term Visibility of Primary Intracerebral Hemorrhage on Magnetic Resonance Imaging Marion Dimigen,
MBChB, MRCS,
Sarah Keir, MRCP, MD, Martin Dennis, and Joanna Wardlaw, FRCP, FRCR, MD
FRCP, MD,
Background: It is unclear whether primary intracerebral hemorrhage (PICH) remains detectable with magnetic resonance imaging (MRI) in the long term, or whether a gradient echo recalled (GRE) sequence is always necessary to detect it. Methods: In a prospectively collected cohort of patients with stroke, we identified survivors of PICH able to undergo MRI at least 3 months after the original PICH. We compared several MRI sequences (spin echo (SE) T2, fast SE (FSE) T2 and proton density, fluid-attenuated inversion recovery, GRE) in a blinded fashion. The number of PICHs visible on each MRI sequence, and the presence of infarcts and microhemorrhages, were determined. Results: In 26 patients imaged 3 years (median) after PICH, between 61% (fluid-attenuated inversion recovery) and 100% (GRE) of PICHs remained identifiable as definite PICH. On FSE T2, 3.4% of PICHs were missed. There were no specific patient features that determined which PICHs remained visible. A new PICH developed in 29% of patients between original presentation and the current study, and 38% had microhemorrhages. Conclusion: Although a FSE T2 sequence will identify most old PICHs, a GRE sequence is essential for definite identification. Recurrent PICH and microhemorrhages appear to be common. Key Words: Magnetic resonance imaging—primary intracerebral hemorrhage. © 2004 by National Stroke Association
Stroke is the third leading cause of death in the developed world.1 Approximately 13% of strokes are caused by spontaneous primary intracerebral hemorrhage (PICH),2 although this may be as few as 6% or as many as 22% according to the data from community-based epidemiology studies.3 Although PICH is associated with severe stroke, some PICHs may be asymptomatic or cause only minor symptoms like transient ischemic attack and the patient may, therefore, not seek medical attention and be diagnosed
From the Department of Clinical Neurosciences, Western General Hospital, Edinburgh, United Kingdom. Received February 9, 2004; accepted March 4, 2004. Supported by NHS R & D Health Technology Assessment Panel grant no: 96/08/01; Chief Scientist Office of the Scottish Office grant no: K/MRS/50/C2457. Address correspondence to: Joanna Wardlaw, FRCP, FRCR, MD, Department of Clinical Neurosciences, Western General Hospital, Crewe Rd, Edinburgh, EH4 2XU United Kingdom. E-mail: jmw@ skull.dcn.ed.ac.uk. 1052-3057/$—see front matter © 2004 by National Stroke Association doi:10.1016/j.jstrokecerebrovasdis.2004.03.002
104
correctly with computed tomography (CT) scanning, acutely.4,5 During the acute stage, hemorrhagic stroke is readily diagnosed by CT scanning, but the distinguishing hyperdensity disappears between 8 days and 3 weeks after the stroke depending on the size of the PICH. After this time, magnetic resonance imaging (MRI) is the investigation of choice.6 Hemoglobin breaks down to leave hemosiderin in the walls of old hematomas, and hemosiderin is thought to persist indefinitely.6,7 Hemosiderin has a characteristic appearance on MRI because of its paramagnetic effects.6 However, it is not known, from pathologic studies, whether all PICHs produce hemosiderin. Nor is it known how reliable the most commonly used MRI sequences, such as fast spin echo (FSE) T2, are in the detection of previous PICH in the subacute or chronic stage. Only one previous study investigated the diagnostic accuracy of different MRI sequences in detecting old brain parenchymal hemorrhage, but this study investigated traumatic intracerebral hemorrhage (106 hematomas in 78 patients).8 One other study followed up old
Journal of Stroke and Cerebrovascular Diseases, Vol. 13, No. 3 (May-June), 2004: pp 104-108
MRI LATE AFTER PICH
PICH but only on CT.9 There have not been any studies that attempted to follow up a cohort of patients with stroke and definite PICH using MRI to determine the proportion remaining detectable as PICH. This knowledge is, however, important as antithrombotic and thrombolytic treatments are increasingly used in the treatment of ischemic stroke. Therefore, this study aimed to determine the accuracy of MRI, and the sensitivity of each commonly used sequence, in diagnosing PICH months to years after the original hemorrhagic stroke.
Materials and Methods All patients with stroke presenting to our hospital were entered into a prospective stroke registry between 1990 and 1999. All underwent routine investigations (including CT), detailed history, and examination by a physician on admission to hospital, and their details were entered into a database. Their stroke type was classified according to the Oxfordshire Community Stroke Project Classification (OCSP) as total anterior circulation stroke, partial anterior circulation stroke, lacunar stroke, or posterior circulation stroke. At the time of the study, 73 patients were identified from the stroke registry as having had a spontaneous PICH, at least 3 months previously, positively diagnosed on CT at presentation, and were still alive at their last follow-up. These patients were contacted through their family doctors and invited to participate in the study. The study was approved by the Lothian Ethics of Medical Research Committee and informed consent was given by all participating patients. The MRI scanning was performed on a 2T scanner (Prestige, GE Medical Systems, Haifa, Israel). Midline sagittal T1, spin echo (SE) T2, FSE T2, FSE proton density (PD), gradient echo recalled (GRE), and fluid-attenuated inversion recovery (FLAIR) sequences were obtained. The sequence parameters were: T1 sagittal–TE 12 milliseconds, TR 500 milliseconds, tip angle 180 degrees; FSE T2 axial–TE 96 milliseconds, TR 5000 milliseconds, tip angle 150 degrees; FSE PD axial–TE 16 milliseconds, TR 2300 milliseconds, tip angle 160 degrees; FLAIR axial–TE 126 milliseconds, TR 6000 milliseconds, TI 2000 milliseconds, tip angle 160 degrees; GRE axial–TE 18 milliseconds, TR 500 milliseconds, tip angle 15 degrees; and SE T2 axial–TE 80 milliseconds, TR 2200 milliseconds, tip angle 180 degrees. The field of view (24) and slice thickness (5 mm) were the same for all sequences, with a 1-mm slice gap for the T1 sagittal, FSE T2, FSE PD, and SE T2; 1.5-mm slice gap for the GRE; and 5-mm slice gap for the FLAIR. The images were printed onto hard copy film (one film/ sequence) for blinded reading. The original CT scans (obtained at the time of the incident stroke) and MRI scans of each patient were reviewed independently by a consultant neuroradiologist
105
and a trainee radiologist. Each MRI sequence was analyzed independently, blind to previous scan results, individual patient details, and clinical outcome. Lesions on the MRI films were noted to be hemorrhage or infarct according to whether or not the low signal of hemosiderin was visible. Other features noted were the presence of microhemorrhages elsewhere in the brain, infarcts with hemorrhagic transformation, enlarged perivascular spaces, white matter hyperintensities, and atrophy. The MRI results were then compared with the original CT scans to determine which of the original hemorrhages were still visible and which were not, and whether any new lesions had appeared.
Results Of the 73 patients identified in the registry as being potentially eligible, about two-thirds did not undergo MRI either because they were disabled, had died since previous follow-up, were claustrophobic, or had intracranial aneurysm clips. In all, 26 patients were able to undergo MRI scanning between 6 and 80 months after the initial stroke (median: 36 months). Their mean age was 66 years (range: 39-83 years) at the time of their original hemorrhage and 69 years (range: 42-87 years) at the time of the MRI scan. Their clinical status at the time of the original PICH, according to the OCSP classification, was: total anterior circulation stroke (3); partial anterior circulation stroke (12); lacunar stroke (2); posterior circulation stroke (5); and 4 patients did not fit clearly into any of these categories (unclassifiable). None of the patients received antithrombotic drugs after the stroke. Of the patients, 3 actually had two PICHs, therefore, 29 separate hemorrhages were identified on the initial CT scan reports. Of a total of 156 MRI scan sequences, 3 individual image sequences were uninterpretable (two SE T2 sequences and one FLAIR sequence) because of patient movement or incomplete imaging as a result of patient intolerance. The results of anomalies seen separately on the CT and MRI scans, for each sequence, are shown in Table 1. On MRI, the number of PICHs detected varied between sequences; on GRE, SE T2, and FSE T2, 38, 36, and 31 hemorrhages, respectively, were seen (i.e., all the original and up to 9 additional hemorrhages, depending on the sequence). On FSE PD and FLAIR, less than the original 29 PICHs were noted as hemorrhages (20 and 23 hemorrhages, respectively). These findings for hemorrhages on the MRI sequences were correlated with the findings on the original CT scans. Table 2 shows the number of the original PICHs, detected as such, and the number of new PICHs visible on MRI, for each MRI sequence. What appeared to be a new PICH developed in up to 29% of patients (depending
M. DIMIGEN ET AL.
106
Table 1. Total number of hemorrhages, infarcts, and other imaging features detected on the original computed tomography scans and later magnetic resonance imaging MRI sequences
Hemorrhage§ Infarct§ Hemorrhagic transformation§ Microhemorrhage‡ Enlarged PVS‡ Small vessel disease‡ Atrophy‡
CT
SE T2*
GRE
FLAIR†
FSE T2
FSE PD
29 8 1 — — — —
36 12 0 3 1 16 18
38 5 2 10 0 8 20
23 12 0 0 1 18 16
31 19 1 2 16 18 10
20 12 0 3 1 16 18
CT, Computed tomography; FLAIR, fluid-attenuated inversion recovery; FSE, fast spin echo; GRE, gradient echo recalled; MRI, magnetic resonance imaging; PD, proton density; PVS, perivascular space; SE, spin echo. *Sequence with two films missing. †Sequence with one film missing. ‡Numbers refer to patients with these features. §Numbers refer to the total number of lesions.
on the MRI sequence) since their presenting stroke. When comparing the CT and MRI side by side, in only one patient was it possible to see an area of cerebromalacia (that had originally been interpreted as an old infarct on CT) that, according to the GRE MRI, had probably actually been an old PICH already present at the time of the presenting PICH. Also, between 8% (FSE T2) and 38% (GRE) of patients showed evidence of microhemorrhages on the MRI scans. A total of 8 cerebral infarcts were noted on the original CT scans in addition to the PICHs. On the MRI sequences, the number varied between 5 (GRE) and 19 (FSE T2). The consultant neuroradiologists’ scan readings were compared with those of the radiology trainee (Table 3). The trainee consistently identified fewer hemorrhages on MRI.
Discussion In this small series of patients surviving to at least 3 months after a stroke because of CT-proven PICH, MRI, using a GRE sequence, detected all their PICHs a median of 3 years (maximum: 6 years and 8 months) later. The other sequences detected fewer old PICHs. In particular, the FSE T2 sequence, one of the most commonly used routine MRI sequences, missed a small percentage of old PICHs. The least sensitive sequence was FSE PD, which made about half of the original PICHs look like infarcts. What are the limitations of the study? We were only able to scan just over one-third of the patients identified in the registry, but in most cases, the failure to scan was because the patient had died or was too disabled to return for imaging. This reflects the high morbidity and mortality of PICH. The patients whom we were able to
Table 2. Correspondence between the original lesion on the computed tomography scan at the time of the incident stroke and the appearance on later magnetic resonance imaging by sequence
Incident PICHs identified on late MRI scans Total No. of extra PICHs seen on MRI presumed to have occurred since original CT No. of new PICHs/patients with new PICH on MRI
SE T2*
GRE
FSE T2
FSE PD
FLAIR†
26/27
29/29
28/29
17/29
17/28
10
9
3
3
6
10/7
9/7
3/3
3/3
6/4
CT, Computed tomography; FLAIR, fluid-attenuated inversion recovery; FSE, fast spin echo; GRE, gradient echo recalled; MRI, magnetic resonance imaging; PD, proton density; PICH, primary intracerebral hemorrhage; SE, spin echo. *Sequence with two films missing. †Sequence with one film missing.
MRI LATE AFTER PICH
107
Table 3. Number of primary intracerebral hemorrhages seen by a radiology trainee compared with a consultant neuroradiologist
SE T2 PICH on both MRI PICH on CT only PICH on MRI only GRE PICH on both MRI PICH on CT only PICH on MRI only FSE PD PICH on both MRI PICH on CT only PICH on MRI only Not clear FSE T2 PICH on both MRI PICH on CT only PICH on MRI only Not clear FLAIR PICH on both MRI PICH on CT only PICH on MRI only
Trainee
Consultant
and CT
24 3 8
26 1 10
and CT
25 4 7
29 0 9
and CT
17 6 6 6
16 12 3
and CT
25 2 9 2
28 1 3
and CT
4 24 1
17 11 6
CT, Computed tomography; FLAIR, fluid-attenuated inversion recovery; FSE, fast spin echo; GRE, gradient echo recalled; MRI, magnetic resonance imaging; PD, proton density; PICH, primary intracerebral hemorrhage; SE, spin echo.
image were more likely to have had a milder original stroke and, therefore, a smaller PICH than those we were unable to image. Therefore, we are unlikely to have overestimated the sensitivity of MRI as larger hemorrhages are less likely to be overlooked on MRI than small ones. Furthermore, able survivors of PICH are more clinically relevant and more representative of patients likely to be seen in clinical practice months to years after a PICH on developing a new illness, the treatment for which might be influenced by knowledge of the patient’s prior PICH. In the patients with more than one PICH, the reason for failure to detect a PICH on some sequences, when other PICHs were visible on that sequence (i.e. failure either to form hemosiderin or for it to be visible) is uncertain. Intracerebral hematomas undergo red cell lysis and break down into ferritin and hemosiderin. These products are taken up by macrophages and astrocytes surrounding the hematoma, eventually forming a hemosiderin-lined slitlike cavity with characteristic appearance on pathology and on MRI,6 and sometimes on CT (although this is less well recognized).9 GRE is the most sensitive sequence for detecting lesions containing paramagnetic material.10,11 GRE also
allows detection of unsuggested regions of hemorrhage for patients with prior head trauma, occult vascular malformations, and hemorrhagic metastases. In this study the GRE sequence picked up the most old and new hemorrhages, and also showed that 38% of patients had microhemorrhages. Previous studies have shown that asymptomatic PICH can occur for patients with no known history of strokelike episodes.4 Patients presenting with minor symptoms may not seek medical attention and may either not be scanned, or be scanned too late to reliably identify PICH on CT.1 Asymptomatic microhemorrhages are found more frequently in patients with hypertension12 and cerebral amyloid angiopathy,13 both risk factors for PICH.14 Hemorrhagic lacunes (small deep cerebral hemorrhages) and silent microhemorrhages may also be markers for increased risk of PICH.15,16 Our observation that about one-third of the patients had other PICHs additional to the one originally diagnosed on CT, and which had occurred in the interval since their incident PICH, is in keeping with other observations17,18 and confirms that recurring PICH is common. It is possible that repeated scanning in our study identified asymptomatic and symptomatic recurrent hemorrhages, as 29% recurrence is higher than the 12% previously documented.18 However, the stroke registry was not set up to collect evidence on recurrent stroke or to know which of the new hemorrhages might have caused symptoms (patients may be sent to other hospitals, the physician may not pass on the information to the stroke registry, patients may forget their symptoms, and it can be difficult to distinguish recurrent stroke from worsening of residual symptoms of prior stroke because of an intercurrent illness). What are the health care implications of this study? Antithrombotic drugs are widely used in primary treatment and secondary prevention of ischemic stroke and for coexisting cardiovascular disease. Aspirin increases the risk of hemorrhagic stroke.19 More specifically, patients being treated for transient ischemic attacks with long-term oral anticoagulants have shown an 8-fold increase in the risk of PICH.17,20 When investigating patients presenting more than 8 to 10 days after stroke (i.e., outside the time frame when a definite diagnosis of hemorrhage could be made with CT–later for more severe strokes),21 a GRE sequence is the most reliable way to detect whether the stroke was PICH. Patients undergoing MRI for investigation of stroke should routinely have a GRE sequence. What are the implications for future research? It is not clear why some PICHs form hemosiderin and others do not. Nor was there any association between any identifiable patient or imaging feature and traumatic hemorrhages that did or did not form hemosiderin.8 There is a lack of data on the frequency and type of recurrent stroke after PICH; current data suggest that the risk of recurrent
108
PICH is frequent, but the data are limited and too imprecise to determine risk factors, including the influence of antithrombotic treatment, reliably.
References 1. Bonita R. Epidemiology of stroke. Lancet 1992;339:342345. 2. Sudlow CLM, Warlow CP. Comparable studies of the incidence of stroke and its pathological types: Results from an international collaboration. Stroke 1997;28:491499. 3. Keir SL, Wardlaw JM, Warlow CP. Stroke epidemiology studies have underestimated the frequency of intracerebral hemorrhage: A systematic review of imaging in epidemiological studies. J Neurol 2002;249:1226-1231. 4. Nakajima Y, Ohsuga H, Yamamoto M, et al. Asymptomatic cerebral hemorrhage detected by MRI. Rinsho Shinkeigaku 1991;31:270-275. 5. Miyashita K, Naritomi H, Nakamura M, et al. Old cerebral hemorrhages in cases of multiple lacunar infarction found by magnetic resonance imaging. Cerebrovasc Dis 1991;1:321-326. 6. Frank JI, Biller J. Laboratory evidence of intracerebral hemorrhage. In: Feldmann E, ed. Intracerebral Hemorrhage. Futura Publishing, 1994:264-267. 7. Garcia JH, Ho KL, Caccamo DV. Intracerebral hemorrhage: Pathology of selected topics. In: Kase CS, Caplan LR, eds. Intracerebral Hemorrhage. Newton, MA: Butterworth-Heinemann, 1994:48-50. 8. Wardlaw JM, Statham PFX. How often is hemosiderin not visible on routine MRI following traumatic intracerebral hemorrhage? Neuroradiology 2000;42:81-84. 9. Franke CL, van Swieten JC, van Gijn J. Residual lesions on computed tomography after intracerebral hemorrhage. Stroke 1991;22:1530-1533. 10. Bradley Jr WG. MR appearance of hemorrhage in the brain. Radiology 1993;189:15-26.
M. DIMIGEN ET AL. 11. Atlas SW, Mark AS, Grossman RI, et al. Intracranial hemorrhage: Gradient echo MR imaging at 1.5T, comparison with spin-echo imaging and clinical applications. Radiology 1988;168:803-807. 12. Roob G, Schmidt R, Kapeller P, et al. MRI evidence of past cerebral microbleeds in a healthy elderly population. Neurology 1999;52:991-994. 13. Greenberg SM, Finklestein SP, Schaefer PW. Petechial hemorrhages accompanying lobar hemorrhage: Detection by gradient-echo MRI. Neurology 1996;46:17511753. 14. Sacco RL, Mayer SA. Epidemiology of intracerebral hemorhage. In: Feldmann E, ed. Intracerebral Hemorrhage. Futura Publishing, 1994:9-15. 15. Scarf J, Brauherr E, Forsting M, et al. Significance of hemorrhagic lacunes on MRI in patients with hypertensive cerebrovascular disease and intracerebral hemorrhage. Neuroradiology 1994;36:504-508. 16. Kidwell CS, Saver JL, Villablanca JP, et al. Magnetic resonance imaging detection of microbleeds before thrombolysis: An emerging application. Stroke 2002;33: 95-98. 17. Bailey RD, Hart RG, Benavente O, et al. Recurrent brain hemorrhage is more frequent than ischemic stroke after intracranial hemorrhage. Neurology 2001;56:773-777. 18. Vermeer SE, Algra A, Franke CL, et al. Long term prognosis after recovery from primary intracerebral hemorrhage. Neurology 2002;59:205-209. 19. He J, Whelton P, Vu B, et al. Aspirin and risk of hemorrhagic stroke: A meta-analysis of randomized controlled trials. JAMA 1998;280:1930-1935. 20. Whisnant JP, Cartlidge NE, Elveback LR. Carotid and vertebrobasilar transient ischemic attacks: Effect of anticoagulants, hypertension and cardiac disorders on survival and stroke occurrence–a population study. Ann Neurol 1978;3:107-115. 21. Wardlaw JM, Keir SL, Dennis MS. The impact of delays in CT brain imaging on the accuracy of diagnosis and subsequent management in patients with minor stroke. J Neurol Neurosurg Psychiatry 2002;74:77-81.
MBChB, MRCS,
Sarah Keir, MRCP, MD, Martin Dennis, and Joanna Wardlaw, FRCP, FRCR, MD
FRCP, MD,
Background: It is unclear whether primary intracerebral hemorrhage (PICH) remains detectable with magnetic resonance imaging (MRI) in the long term, or whether a gradient echo recalled (GRE) sequence is always necessary to detect it. Methods: In a prospectively collected cohort of patients with stroke, we identified survivors of PICH able to undergo MRI at least 3 months after the original PICH. We compared several MRI sequences (spin echo (SE) T2, fast SE (FSE) T2 and proton density, fluid-attenuated inversion recovery, GRE) in a blinded fashion. The number of PICHs visible on each MRI sequence, and the presence of infarcts and microhemorrhages, were determined. Results: In 26 patients imaged 3 years (median) after PICH, between 61% (fluid-attenuated inversion recovery) and 100% (GRE) of PICHs remained identifiable as definite PICH. On FSE T2, 3.4% of PICHs were missed. There were no specific patient features that determined which PICHs remained visible. A new PICH developed in 29% of patients between original presentation and the current study, and 38% had microhemorrhages. Conclusion: Although a FSE T2 sequence will identify most old PICHs, a GRE sequence is essential for definite identification. Recurrent PICH and microhemorrhages appear to be common. Key Words: Magnetic resonance imaging—primary intracerebral hemorrhage. © 2004 by National Stroke Association
Stroke is the third leading cause of death in the developed world.1 Approximately 13% of strokes are caused by spontaneous primary intracerebral hemorrhage (PICH),2 although this may be as few as 6% or as many as 22% according to the data from community-based epidemiology studies.3 Although PICH is associated with severe stroke, some PICHs may be asymptomatic or cause only minor symptoms like transient ischemic attack and the patient may, therefore, not seek medical attention and be diagnosed
From the Department of Clinical Neurosciences, Western General Hospital, Edinburgh, United Kingdom. Received February 9, 2004; accepted March 4, 2004. Supported by NHS R & D Health Technology Assessment Panel grant no: 96/08/01; Chief Scientist Office of the Scottish Office grant no: K/MRS/50/C2457. Address correspondence to: Joanna Wardlaw, FRCP, FRCR, MD, Department of Clinical Neurosciences, Western General Hospital, Crewe Rd, Edinburgh, EH4 2XU United Kingdom. E-mail: jmw@ skull.dcn.ed.ac.uk. 1052-3057/$—see front matter © 2004 by National Stroke Association doi:10.1016/j.jstrokecerebrovasdis.2004.03.002
104
correctly with computed tomography (CT) scanning, acutely.4,5 During the acute stage, hemorrhagic stroke is readily diagnosed by CT scanning, but the distinguishing hyperdensity disappears between 8 days and 3 weeks after the stroke depending on the size of the PICH. After this time, magnetic resonance imaging (MRI) is the investigation of choice.6 Hemoglobin breaks down to leave hemosiderin in the walls of old hematomas, and hemosiderin is thought to persist indefinitely.6,7 Hemosiderin has a characteristic appearance on MRI because of its paramagnetic effects.6 However, it is not known, from pathologic studies, whether all PICHs produce hemosiderin. Nor is it known how reliable the most commonly used MRI sequences, such as fast spin echo (FSE) T2, are in the detection of previous PICH in the subacute or chronic stage. Only one previous study investigated the diagnostic accuracy of different MRI sequences in detecting old brain parenchymal hemorrhage, but this study investigated traumatic intracerebral hemorrhage (106 hematomas in 78 patients).8 One other study followed up old
Journal of Stroke and Cerebrovascular Diseases, Vol. 13, No. 3 (May-June), 2004: pp 104-108
MRI LATE AFTER PICH
PICH but only on CT.9 There have not been any studies that attempted to follow up a cohort of patients with stroke and definite PICH using MRI to determine the proportion remaining detectable as PICH. This knowledge is, however, important as antithrombotic and thrombolytic treatments are increasingly used in the treatment of ischemic stroke. Therefore, this study aimed to determine the accuracy of MRI, and the sensitivity of each commonly used sequence, in diagnosing PICH months to years after the original hemorrhagic stroke.
Materials and Methods All patients with stroke presenting to our hospital were entered into a prospective stroke registry between 1990 and 1999. All underwent routine investigations (including CT), detailed history, and examination by a physician on admission to hospital, and their details were entered into a database. Their stroke type was classified according to the Oxfordshire Community Stroke Project Classification (OCSP) as total anterior circulation stroke, partial anterior circulation stroke, lacunar stroke, or posterior circulation stroke. At the time of the study, 73 patients were identified from the stroke registry as having had a spontaneous PICH, at least 3 months previously, positively diagnosed on CT at presentation, and were still alive at their last follow-up. These patients were contacted through their family doctors and invited to participate in the study. The study was approved by the Lothian Ethics of Medical Research Committee and informed consent was given by all participating patients. The MRI scanning was performed on a 2T scanner (Prestige, GE Medical Systems, Haifa, Israel). Midline sagittal T1, spin echo (SE) T2, FSE T2, FSE proton density (PD), gradient echo recalled (GRE), and fluid-attenuated inversion recovery (FLAIR) sequences were obtained. The sequence parameters were: T1 sagittal–TE 12 milliseconds, TR 500 milliseconds, tip angle 180 degrees; FSE T2 axial–TE 96 milliseconds, TR 5000 milliseconds, tip angle 150 degrees; FSE PD axial–TE 16 milliseconds, TR 2300 milliseconds, tip angle 160 degrees; FLAIR axial–TE 126 milliseconds, TR 6000 milliseconds, TI 2000 milliseconds, tip angle 160 degrees; GRE axial–TE 18 milliseconds, TR 500 milliseconds, tip angle 15 degrees; and SE T2 axial–TE 80 milliseconds, TR 2200 milliseconds, tip angle 180 degrees. The field of view (24) and slice thickness (5 mm) were the same for all sequences, with a 1-mm slice gap for the T1 sagittal, FSE T2, FSE PD, and SE T2; 1.5-mm slice gap for the GRE; and 5-mm slice gap for the FLAIR. The images were printed onto hard copy film (one film/ sequence) for blinded reading. The original CT scans (obtained at the time of the incident stroke) and MRI scans of each patient were reviewed independently by a consultant neuroradiologist
105
and a trainee radiologist. Each MRI sequence was analyzed independently, blind to previous scan results, individual patient details, and clinical outcome. Lesions on the MRI films were noted to be hemorrhage or infarct according to whether or not the low signal of hemosiderin was visible. Other features noted were the presence of microhemorrhages elsewhere in the brain, infarcts with hemorrhagic transformation, enlarged perivascular spaces, white matter hyperintensities, and atrophy. The MRI results were then compared with the original CT scans to determine which of the original hemorrhages were still visible and which were not, and whether any new lesions had appeared.
Results Of the 73 patients identified in the registry as being potentially eligible, about two-thirds did not undergo MRI either because they were disabled, had died since previous follow-up, were claustrophobic, or had intracranial aneurysm clips. In all, 26 patients were able to undergo MRI scanning between 6 and 80 months after the initial stroke (median: 36 months). Their mean age was 66 years (range: 39-83 years) at the time of their original hemorrhage and 69 years (range: 42-87 years) at the time of the MRI scan. Their clinical status at the time of the original PICH, according to the OCSP classification, was: total anterior circulation stroke (3); partial anterior circulation stroke (12); lacunar stroke (2); posterior circulation stroke (5); and 4 patients did not fit clearly into any of these categories (unclassifiable). None of the patients received antithrombotic drugs after the stroke. Of the patients, 3 actually had two PICHs, therefore, 29 separate hemorrhages were identified on the initial CT scan reports. Of a total of 156 MRI scan sequences, 3 individual image sequences were uninterpretable (two SE T2 sequences and one FLAIR sequence) because of patient movement or incomplete imaging as a result of patient intolerance. The results of anomalies seen separately on the CT and MRI scans, for each sequence, are shown in Table 1. On MRI, the number of PICHs detected varied between sequences; on GRE, SE T2, and FSE T2, 38, 36, and 31 hemorrhages, respectively, were seen (i.e., all the original and up to 9 additional hemorrhages, depending on the sequence). On FSE PD and FLAIR, less than the original 29 PICHs were noted as hemorrhages (20 and 23 hemorrhages, respectively). These findings for hemorrhages on the MRI sequences were correlated with the findings on the original CT scans. Table 2 shows the number of the original PICHs, detected as such, and the number of new PICHs visible on MRI, for each MRI sequence. What appeared to be a new PICH developed in up to 29% of patients (depending
M. DIMIGEN ET AL.
106
Table 1. Total number of hemorrhages, infarcts, and other imaging features detected on the original computed tomography scans and later magnetic resonance imaging MRI sequences
Hemorrhage§ Infarct§ Hemorrhagic transformation§ Microhemorrhage‡ Enlarged PVS‡ Small vessel disease‡ Atrophy‡
CT
SE T2*
GRE
FLAIR†
FSE T2
FSE PD
29 8 1 — — — —
36 12 0 3 1 16 18
38 5 2 10 0 8 20
23 12 0 0 1 18 16
31 19 1 2 16 18 10
20 12 0 3 1 16 18
CT, Computed tomography; FLAIR, fluid-attenuated inversion recovery; FSE, fast spin echo; GRE, gradient echo recalled; MRI, magnetic resonance imaging; PD, proton density; PVS, perivascular space; SE, spin echo. *Sequence with two films missing. †Sequence with one film missing. ‡Numbers refer to patients with these features. §Numbers refer to the total number of lesions.
on the MRI sequence) since their presenting stroke. When comparing the CT and MRI side by side, in only one patient was it possible to see an area of cerebromalacia (that had originally been interpreted as an old infarct on CT) that, according to the GRE MRI, had probably actually been an old PICH already present at the time of the presenting PICH. Also, between 8% (FSE T2) and 38% (GRE) of patients showed evidence of microhemorrhages on the MRI scans. A total of 8 cerebral infarcts were noted on the original CT scans in addition to the PICHs. On the MRI sequences, the number varied between 5 (GRE) and 19 (FSE T2). The consultant neuroradiologists’ scan readings were compared with those of the radiology trainee (Table 3). The trainee consistently identified fewer hemorrhages on MRI.
Discussion In this small series of patients surviving to at least 3 months after a stroke because of CT-proven PICH, MRI, using a GRE sequence, detected all their PICHs a median of 3 years (maximum: 6 years and 8 months) later. The other sequences detected fewer old PICHs. In particular, the FSE T2 sequence, one of the most commonly used routine MRI sequences, missed a small percentage of old PICHs. The least sensitive sequence was FSE PD, which made about half of the original PICHs look like infarcts. What are the limitations of the study? We were only able to scan just over one-third of the patients identified in the registry, but in most cases, the failure to scan was because the patient had died or was too disabled to return for imaging. This reflects the high morbidity and mortality of PICH. The patients whom we were able to
Table 2. Correspondence between the original lesion on the computed tomography scan at the time of the incident stroke and the appearance on later magnetic resonance imaging by sequence
Incident PICHs identified on late MRI scans Total No. of extra PICHs seen on MRI presumed to have occurred since original CT No. of new PICHs/patients with new PICH on MRI
SE T2*
GRE
FSE T2
FSE PD
FLAIR†
26/27
29/29
28/29
17/29
17/28
10
9
3
3
6
10/7
9/7
3/3
3/3
6/4
CT, Computed tomography; FLAIR, fluid-attenuated inversion recovery; FSE, fast spin echo; GRE, gradient echo recalled; MRI, magnetic resonance imaging; PD, proton density; PICH, primary intracerebral hemorrhage; SE, spin echo. *Sequence with two films missing. †Sequence with one film missing.
MRI LATE AFTER PICH
107
Table 3. Number of primary intracerebral hemorrhages seen by a radiology trainee compared with a consultant neuroradiologist
SE T2 PICH on both MRI PICH on CT only PICH on MRI only GRE PICH on both MRI PICH on CT only PICH on MRI only FSE PD PICH on both MRI PICH on CT only PICH on MRI only Not clear FSE T2 PICH on both MRI PICH on CT only PICH on MRI only Not clear FLAIR PICH on both MRI PICH on CT only PICH on MRI only
Trainee
Consultant
and CT
24 3 8
26 1 10
and CT
25 4 7
29 0 9
and CT
17 6 6 6
16 12 3
and CT
25 2 9 2
28 1 3
and CT
4 24 1
17 11 6
CT, Computed tomography; FLAIR, fluid-attenuated inversion recovery; FSE, fast spin echo; GRE, gradient echo recalled; MRI, magnetic resonance imaging; PD, proton density; PICH, primary intracerebral hemorrhage; SE, spin echo.
image were more likely to have had a milder original stroke and, therefore, a smaller PICH than those we were unable to image. Therefore, we are unlikely to have overestimated the sensitivity of MRI as larger hemorrhages are less likely to be overlooked on MRI than small ones. Furthermore, able survivors of PICH are more clinically relevant and more representative of patients likely to be seen in clinical practice months to years after a PICH on developing a new illness, the treatment for which might be influenced by knowledge of the patient’s prior PICH. In the patients with more than one PICH, the reason for failure to detect a PICH on some sequences, when other PICHs were visible on that sequence (i.e. failure either to form hemosiderin or for it to be visible) is uncertain. Intracerebral hematomas undergo red cell lysis and break down into ferritin and hemosiderin. These products are taken up by macrophages and astrocytes surrounding the hematoma, eventually forming a hemosiderin-lined slitlike cavity with characteristic appearance on pathology and on MRI,6 and sometimes on CT (although this is less well recognized).9 GRE is the most sensitive sequence for detecting lesions containing paramagnetic material.10,11 GRE also
allows detection of unsuggested regions of hemorrhage for patients with prior head trauma, occult vascular malformations, and hemorrhagic metastases. In this study the GRE sequence picked up the most old and new hemorrhages, and also showed that 38% of patients had microhemorrhages. Previous studies have shown that asymptomatic PICH can occur for patients with no known history of strokelike episodes.4 Patients presenting with minor symptoms may not seek medical attention and may either not be scanned, or be scanned too late to reliably identify PICH on CT.1 Asymptomatic microhemorrhages are found more frequently in patients with hypertension12 and cerebral amyloid angiopathy,13 both risk factors for PICH.14 Hemorrhagic lacunes (small deep cerebral hemorrhages) and silent microhemorrhages may also be markers for increased risk of PICH.15,16 Our observation that about one-third of the patients had other PICHs additional to the one originally diagnosed on CT, and which had occurred in the interval since their incident PICH, is in keeping with other observations17,18 and confirms that recurring PICH is common. It is possible that repeated scanning in our study identified asymptomatic and symptomatic recurrent hemorrhages, as 29% recurrence is higher than the 12% previously documented.18 However, the stroke registry was not set up to collect evidence on recurrent stroke or to know which of the new hemorrhages might have caused symptoms (patients may be sent to other hospitals, the physician may not pass on the information to the stroke registry, patients may forget their symptoms, and it can be difficult to distinguish recurrent stroke from worsening of residual symptoms of prior stroke because of an intercurrent illness). What are the health care implications of this study? Antithrombotic drugs are widely used in primary treatment and secondary prevention of ischemic stroke and for coexisting cardiovascular disease. Aspirin increases the risk of hemorrhagic stroke.19 More specifically, patients being treated for transient ischemic attacks with long-term oral anticoagulants have shown an 8-fold increase in the risk of PICH.17,20 When investigating patients presenting more than 8 to 10 days after stroke (i.e., outside the time frame when a definite diagnosis of hemorrhage could be made with CT–later for more severe strokes),21 a GRE sequence is the most reliable way to detect whether the stroke was PICH. Patients undergoing MRI for investigation of stroke should routinely have a GRE sequence. What are the implications for future research? It is not clear why some PICHs form hemosiderin and others do not. Nor was there any association between any identifiable patient or imaging feature and traumatic hemorrhages that did or did not form hemosiderin.8 There is a lack of data on the frequency and type of recurrent stroke after PICH; current data suggest that the risk of recurrent
108
PICH is frequent, but the data are limited and too imprecise to determine risk factors, including the influence of antithrombotic treatment, reliably.
References 1. Bonita R. Epidemiology of stroke. Lancet 1992;339:342345. 2. Sudlow CLM, Warlow CP. Comparable studies of the incidence of stroke and its pathological types: Results from an international collaboration. Stroke 1997;28:491499. 3. Keir SL, Wardlaw JM, Warlow CP. Stroke epidemiology studies have underestimated the frequency of intracerebral hemorrhage: A systematic review of imaging in epidemiological studies. J Neurol 2002;249:1226-1231. 4. Nakajima Y, Ohsuga H, Yamamoto M, et al. Asymptomatic cerebral hemorrhage detected by MRI. Rinsho Shinkeigaku 1991;31:270-275. 5. Miyashita K, Naritomi H, Nakamura M, et al. Old cerebral hemorrhages in cases of multiple lacunar infarction found by magnetic resonance imaging. Cerebrovasc Dis 1991;1:321-326. 6. Frank JI, Biller J. Laboratory evidence of intracerebral hemorrhage. In: Feldmann E, ed. Intracerebral Hemorrhage. Futura Publishing, 1994:264-267. 7. Garcia JH, Ho KL, Caccamo DV. Intracerebral hemorrhage: Pathology of selected topics. In: Kase CS, Caplan LR, eds. Intracerebral Hemorrhage. Newton, MA: Butterworth-Heinemann, 1994:48-50. 8. Wardlaw JM, Statham PFX. How often is hemosiderin not visible on routine MRI following traumatic intracerebral hemorrhage? Neuroradiology 2000;42:81-84. 9. Franke CL, van Swieten JC, van Gijn J. Residual lesions on computed tomography after intracerebral hemorrhage. Stroke 1991;22:1530-1533. 10. Bradley Jr WG. MR appearance of hemorrhage in the brain. Radiology 1993;189:15-26.
M. DIMIGEN ET AL. 11. Atlas SW, Mark AS, Grossman RI, et al. Intracranial hemorrhage: Gradient echo MR imaging at 1.5T, comparison with spin-echo imaging and clinical applications. Radiology 1988;168:803-807. 12. Roob G, Schmidt R, Kapeller P, et al. MRI evidence of past cerebral microbleeds in a healthy elderly population. Neurology 1999;52:991-994. 13. Greenberg SM, Finklestein SP, Schaefer PW. Petechial hemorrhages accompanying lobar hemorrhage: Detection by gradient-echo MRI. Neurology 1996;46:17511753. 14. Sacco RL, Mayer SA. Epidemiology of intracerebral hemorhage. In: Feldmann E, ed. Intracerebral Hemorrhage. Futura Publishing, 1994:9-15. 15. Scarf J, Brauherr E, Forsting M, et al. Significance of hemorrhagic lacunes on MRI in patients with hypertensive cerebrovascular disease and intracerebral hemorrhage. Neuroradiology 1994;36:504-508. 16. Kidwell CS, Saver JL, Villablanca JP, et al. Magnetic resonance imaging detection of microbleeds before thrombolysis: An emerging application. Stroke 2002;33: 95-98. 17. Bailey RD, Hart RG, Benavente O, et al. Recurrent brain hemorrhage is more frequent than ischemic stroke after intracranial hemorrhage. Neurology 2001;56:773-777. 18. Vermeer SE, Algra A, Franke CL, et al. Long term prognosis after recovery from primary intracerebral hemorrhage. Neurology 2002;59:205-209. 19. He J, Whelton P, Vu B, et al. Aspirin and risk of hemorrhagic stroke: A meta-analysis of randomized controlled trials. JAMA 1998;280:1930-1935. 20. Whisnant JP, Cartlidge NE, Elveback LR. Carotid and vertebrobasilar transient ischemic attacks: Effect of anticoagulants, hypertension and cardiac disorders on survival and stroke occurrence–a population study. Ann Neurol 1978;3:107-115. 21. Wardlaw JM, Keir SL, Dennis MS. The impact of delays in CT brain imaging on the accuracy of diagnosis and subsequent management in patients with minor stroke. J Neurol Neurosurg Psychiatry 2002;74:77-81.