Evaluation of Traumatic Subarachnoid Hemorrhage Using Susceptibility-Weighted ORIGINAL RESEARCH Imaging Z. Wu BACKGROUND AND PURPOSE: SWI is an MR imaging technique that is very sensitive to hemorrhage. S. Li Our goal was to compare SWI and CT to determine if SWI can show traumatic SAH in different parts of the subarachnoid space. J. Lei D. An MATERIALS AND METHODS: Twenty acute TBI patients identified by CT with SAH underwent MR E.M. Haacke imaging scans. Two neuroradiologists analyzed the CT and SWI data to decide whether there were SAHs in 8 anatomical parts of the subarachnoid space. RESULTS: Fifty-five areas with SAH were identified by both CT and SWI. Ten areas were identified by CT only and 13 by SWI only. SAH was recognized on SWI by its very dark signal intensity surrounded by CSF signal intensity in the sulci or cisterns. Compared with the smooth-looking veins, SAH tended to have a rough boundary and inhomogeneous signal intensity. In many instances, blood in the sulcus left an area of signal intensity loss that had a “triangle” shape. SWI showed 5 more cases of intraventricular hemorrhage than did CT. CONCLUSIONS: SAH can be recognized by SWI through its signal intensity and unique morphology. SWI can provide complementary information to CT in terms of small amounts of SAH and hemorrhage inside the ventricles. ABBREVIATIONS: BW ϭ bandwidth; FA ϭ flip angle; FLAIR ϭ fluid-attenuated inversion recovery; FPC ϭ frontal-parietal convexity; GRAPPA ϭ generalized autocalibrating partially parallel acquisition; HU ϭ Hounsfield unit; IHF ϭ interhemispheric fissure; IVH ϭ intraventricular hemorrhage; kvp ϭ kilovoltage peak; MIP ϭ minimum intensity projection; PFC ϭ posterior fossa cisterns; PMC ϭ perimesencephalic cistern; SAH ϭ subarachnoid hemorrhage; SVF ϭ Sylvian fissure; SWI ϭ susceptibility-weighted imaging; T1WI ϭ T1-weighted imaging; T2WI ϭ T2-weighted imaging; TBI ϭ traumatic brain injury; TNC ϭ tentorial cistern; TOC ϭ temporal-occipital convexity; WM ϭ white matter ubarachnoid hemorrhage is a common finding in the set- CT has long been considered the first choice in the detec- Sting of acute TBI.1 Traumatic injury to the brain results in tion of acute SAH because of its sensitivity, low cost, and wide stretching, tearing, and laceration of the blood vessels cours- availability. The sensitivity of CT in detecting SAH is depen- ing within the subarachnoid space where blood enters the sub- dent on the resolution of the scanner, the interval after onset, arachnoid space and mixes with the CSF within the subarach- the amount of hemorrhage, and the skills of the radiologist.8 It noid space.2 Hemorrhage can accumulate both in the brain has been widely accepted by neuroradiologists that MR imag- cisterns and fissures as well as in the sulci throughout the brain ing is insensitive to acute SAH, and the reason why SAH can- 3 after injury. Intraventricular hemorrhage can also be seen not be seen with MR imaging consistently has been attributed along with SAH, as the CSF spaces are communicating com- to the complex hemorrhagic signal intensity seen on MR im- partments. The incidence of traumatic SAH varies from 11% aging.9-11 During the past 15 years, new MR imaging se- 4-7 to 60% in TBI patients. A few large TBI studies have shown quences have been developed and their application in detect- that the amount of SAH significantly correlates with initial ing SAH at high fields has been explored. Some studies have 4-6 TBI presentation and long-term outcomes. shown that FLAIR is more sensitive than CT in detecting SAH in the acute stage in vitro and in vivo.12,13 However, hyperin- Received July 22, 2009; accepted after revision November 23. tensity in the subarachnoid space on FLAIR can be caused by From the School for Biomedical Engineering (Z.W., E.M.H.), McMaster University, Hamilton, artifacts such as supplemental oxygen, CSF pulsation, and vas- Ontario, Canada; Department of Radiology (S.L., D.a.), Capital Medical University, Beijing 14 Tiantan Hospital, Beijing, China; Department of Radiology (J.L.), Tianjin Huan Hu Hospital, cular pulsation rather than SAH. On the other hand, gradi- Tianjin, China; Department of Radiology (E.M.H.), Wayne State University, Detroit, Mich- ent-echo sequences have been used in imaging hyperacute igan; and The MRI Institute for Biomedical Research (E.M.H.), Detroit, Michigan. SAH in a small number of patients and some think that it can This work was supported in part by National Institutes of Health Award NHLBI be reliably used to diagnose SAH at 3T.15 R01HL062983. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Heart, Lung, and Blood Institute or SWI is an MR imaging technique that uses tissue magnetic 16 the National Institutes of Health. susceptibility differences to generate a unique contrast. SWI Please address correspondence to Li Shaowu, Neuroimaging Center, Beijing Tiantan data are composed of 4 sets of images: the magnitude, the Hospital, 6 Tiantan Xili, Chongwen District, Beijing, China 100050; e-mail: [email protected] phase, the final processed SWI, and the MIP (often over 4 adjacent sections) images. The original magnitude image is a Indicates open access to non-subscribers at www.ajnr.org flow-compensated gradient-echo image but with a long TE. DOI 10.3174/ajnr.A2022 Usually TE is 40 ms at 1.5T and 20 ms at 3T for SWI. These 1302 Wu ͉ AJNR 31 ͉ Aug 2010 ͉ www.ajnr.org values render the phase images at both field strengths essen- 0.5 ϫ 5mm3, and FOV ϭ 240 mm; from Beijing Tiantan Hospital the tially identical because the phase is proportional to the prod- parameters were: x-ray tube current ϭ 310 mA, kvp ϭ 120 kV, con- uct of the field strength and echo time. The magnitude image volution kernel H31s (soft tissue window), resolution ϭ 0.5 ϫ 0.5 ϫ in a gradient-echo scan tends to highlight small changes in 4.5 mm3, and FOV ϭ 230 mm. susceptibility across a voxel as signal intensity loss.17 The SWI phase image is high pass filtered to remove low spatial fre- MR Imaging Scan Parameters quency background field variations. This image is then used to Four MR imaging sequences were collected at 3T (Trio; Siemens), create a mask to highlight the remaining information coming including: T1WI, T2WI, FLAIR, and SWI. The SWI parameters from from smaller structures, including veins, hemorrhage, and an- Tianjin Huanhu Hospital were: TR/TE ϭ 29/20 ms, FA ϭ 15°, BW ϭ atomic structures that contain more iron such as red nucleus 120 Hz/pixel, spatial resolution ϭ 0.5 ϫ 0.5 ϫ 2.0 mm3, FOV ϭ 256 and substantia nigra, and calcium deposits.18 The SWI-filtered mm, parallel imaging mode ϭ GRAPPA, accelerating factor ϭ 2; phase images themselves are also useful. For example, iron and from Beijing Tiantan Hospital the parameters were: TR/TE ϭ 28/20 calcium can be differentiated from each other because iron is ms, FA ϭ 15°, BW ϭ 120 Hz/pixel, spatial resolution ϭ 0.5 ϫ 0.5 ϫ paramagnetic whereas calcium is diamagnetic. Therefore, 2.0 mm3, FOV ϭ 230 mm, parallel imaging mode ϭ GRAPPA, accel- similar signals on magnitude images will have opposite phase erating factor ϭ 2. values (with iron in veins appearing dark while calcium depos- its often appear bright for a right-handed system).18 Because Data Evaluation phase values range from Ϫ␲ to ␲, aliasing may occur in areas Conventionally, subarachnoid space is divided into the following ma- having very high iron or calcium content. When phase aliases, jor parts: 1) convexity subarachnoid spaces, which cover the surfaces it changes from dark to bright abruptly and creates a zebra of the brain’s hemispheres; 2) Sylvian cistern, which is between the stripe pattern that is easily recognized. This aliasing creates a insula and opercula; 3) mesencephalic cisterns, which refer to the area heterogeneous, discontinuous phase signal intensity. Aliasing around midbrain, 4) suprasellar (basal) cisterns, which lie above the also makes it difficult to quantify the amount of iron, though sella and between the cerebral peduncles; and 5) posterior fossa cis- in some cases it is possible to unwrap the phase aliasing by terns, which include the medullary cistern, cisterna magna, pontine using postprocessing methods.19 cistern, cerebellopontine angle cistern, and superior cerebellar SWI has been successfully applied in TBI and has proved to cistern.21 be from 3 to 6 times more sensitive than conventional gradi- In this study, we did not strictly follow this conventional catego- ent-echo imaging in detecting parenchymal hemorrhage.20 rizing method, but rather we divided or combined the above 5 major No study has yet been done by using SWI to evaluate SAH. In parts into 8 regions as described below. To start, we divided the con- this study, we compare CT and SWI in their abilities to detect vexity subarachnoid spaces into 3 parts: 1) frontal-parietal, 2) tempo- BRAIN SAH and determine whether SWI can provide complementary ral-occipital, and 3) interhemispheric cisterns; followed by 4) the Syl- information to CT. vian cistern, 5) the perimesencephalic cisterns (combination of mesencephalic cisterns and basal cisterns), 6) posterior fossa cisterns, 7) tentorial cistern (it is also called superior cerebella cistern and Materials and Methods ORIGINAL RESEARCH belongs to the posterior fossa cisterns, but it was evaluated separately Patient Data Collection in our study), and 8) intraventricular hemorrhage. Two neuroradi- From May 2008 to December 2008, patients from the Emergency ologists were blinded to all other information except the imaging Department or Trauma Center from Tianjin Huanhu Hospital and data.