Cerebral vasospasm remains a major cause of morbidity and mortality among patients after subarachnoid hemorrhage (SAH) from ruptured intracranial aneurysms. Characteristically occurring 3–14 days after initial hemorrhage, vasospasm is the narrowing of the intracranial cerebral arteries that may cause ischemia and potentially infarction. With an estimated incidence of 6–16 per 100 000, aneurysm rupture is the etiology in approximately 5–15% of stroke patients, with overall high mortality ranging from 30–70% . Approximately 35% die before arrival at a hospital or within the first 24 h [1–3]. Those surviving the initial hemorrhage, and who subsequently undergo surgical clipping or endovascular coiling, will experience the greatest morbidity and mortality as a result of cerebral vasospasm. About 30–70% of patients with aneurysmal SAH will develop cerebral angiographic vasospasm, with delayed neurologic deficits manifesting in 30–50% and permanent disability in 12–17% [1–4].
During the first days and weeks after hemorrhage, patients are closely monitored within a neurological intensive care unit with strict maintenance of cerebral perfusion pressures to compensate for aberrations in cerebral autoregulation and to ultimately prevent vasospasm induced ischemia. This period necessitates close working relationships between radiology and neurocritical care personnel, to ensure that diagnosis and treatment are carried out expeditiously. In addition to clinical examination, patients are often evaluated daily by transcranial Doppler (TCD) ultrasound and periodically by other noninvasive modalities such as single photon emission computed tomography (SPECT) or perfusion CT [4–7]. The use of SPECT has been well established as an effective method to evaluate cerebral perfusion and, in conjunction with clinical exam and TCD ultrasound, can provide a complement to predict those patients affected with cerebral vasospasm [8–13]. To date, SPECT images are qualitatively evaluated often in comparison with recent prior studies, though automated software now exists to quantify perfusion changes in an operator-independent manner. Quantitative software programs are currently used in nuclear medicine evaluation of cardiac perfusion as well as cerebral blood flow. Several different software programs are available for quantification of perfusion abnormalities. Hermes brain registration and analysis software (BRASS, Hermes Medical Solutions, Stockholm, Sweden) is a state-of-the-art, quantitative analysis program for automatic fitting of brain perfusion scans to quantify and localize regions of abnormal perfusion [14–16]. This software is used for a variety of pathological conditions, including Alzheimer's disease and other dementias, Parkinson's disease and other essential tremors, and cerebro-vascular disorders like stroke and transient ischemic attacks [17–21]. Our aim was to evaluate the utility of semiquantitative SPECT using Hermes BRASS software relative to TCD ultrasound and qualitative SPECT interpretation within a cohort of patients, all with cerebral vasospasm as measured by digital subtraction angiography (DSA).
With the approval from the institutional review board, retrospective analysis of all patients admitted for aneurysmal SAH between January 2005 and June 2007 to Harborview Medical Center (n = 301) was performed with subselection of all patients undergoing intra-arterial treatment for cerebral vasospasm (n = 73). A subset of 32 nonconsecutive patients, selected for completeness of data (18 male and 14 female; age 20–71 years), was included in the study and analyzed for perfusion deficits in terms of anterior, middle and posterior vascular distributions – using both TCD ultrasonography and brain perfusion SPECT studies.
All patients were admitted by the neurosurgical service and managed within the NICU using standard hemodilution, hypertension, and hypervolemia therapy. All patients were subjected to a baseline SPECT study within a few hours of initial treatment of bleeding aneurysm either with coiling or clipping. All patients underwent daily clinical neurological evaluation and TCDs to assess the development of vasospasm. If patients showed neurological deterioration on clinical examination, or evidence of vasospasm by TCD, brain perfusion SPECT (preintervention) examination was performed for additional evaluation before patients were scheduled for neurological interventional procedure. A patient underwent DSA if they met three of the following four criteria: (i) new onset neurological deficit without other known cause; (ii) persistent neurological deficit despite optimized triple ‘H’ management; (iii) new or worsening vasospasm by TCD; and/or (iv) new or worsening hypoperfusion by SPECT.
TCD ultrasounds were performed by certified technologists and interpreted by a neurocritical care physician. Lindegaard ratios were calculated for middle cerebral artery (MCA) and basilar arteries to determine vasospasm while velocities (<120, 120–200, >200 cm/s) for the anterior cerebral artery (ACA) and posterior cerebral artery were used to determine mild, moderate, and severe vasospasm.
Brain SPECT and BRASS analysis
Each patient was injected with approximately 1110 MBq (30 mCi) 99mTc ethyl cysteinate dimer and images were obtained approximately 35 min later. All images were acquired using a triple-headed tomographic scanner and a low-energy, high-resolution collimator. A 20% window was centered on the 140 keV photopeak of 99mTc. SPECT images were acquired in a step-and-shoot manner with 64 steps, each lasting 25 s, acquired over 360° using clockwise rotation. Images were processed with a Wiener prefilter and ramp filter for resolution recovery. Software attenuation correction with a coefficient of 0.11/cm was used in all patients with intact cranial bones. All images were reconstructed in transaxial, coronal, and sagittal planes that were approximately 6 mm thick.
Baseline and preintervention SPECT studies were analyzed qualitatively by a qualified nuclear medicine physician (DL), experienced in reading brain perfusion studies. Qualitative evaluation was performed by comparing the contralateral hemispheric and cerebellar uptake of the radiotracer and also comparing the baseline studies, which are obtained shortly after hospital admission. Quantitative evaluation using BRASS software (Hermes Medical Solutions) was then performed comparing the baseline study with the prevascular intervention study. These studies were evaluated for interval decrease in perfusion, resulting from vasospasm. BRASS software automatically fits and quantifies brain perfusion scans. This software can either compare a study to three-dimensional (3D) reference templates (created from images of normal patients) with automatic masking or it can compare two studies with each other and quantify the differences. If two studies are loaded, then after fitting, both will be aligned with each other and the template. The process requires the transverse slice sets of 1 or 2 studies for automatic alignment with the normal template. The software performs normalization using total counts, maximum counts, or region of interest depending on the operator input in terms of standard deviation and difference thresholds for identification of perfusion defects. After all the processing is done, the software displays 1–3 views simultaneously with reference images and cross hairs to mark points simultaneously on all three views which can be scrolled through easily. Regional map quantification is performed using voxel-based and volume region-based quantification techniques and provides with region-based statistics of activity in each region in comparison with the template. Percentage difference between the two scans is also provided if comparing two studies on the same patient performed at different times. The perfusion defect is quantified with respect to defect size and severity and the number of defective voxels. The other advantage of the software is that the resulting defects can be overlaid on anatomical images – either CT or MRI for accurate localization of the defects anatomically. For this study, normalization to global counts was used.
A difference threshold of 12% with global counts normalization was chosen in calculating the differences in perfusion. This threshold was chosen because it has been shown in previous studies that a variation of about 5% in two resting scans separated by 48 h has been described in literature . The results were graded from 0 to 3 based on the percentage numbers obtained after BRASS analysis to enable comparison with angiographic spasm, which was also graded from 0 to 3 (0 – normal, 1 – mild, 2 – moderate, and 3 – severe; mild – 12–19%, moderate – 20–29%, and severe >30%).
Data collection was compiled in a spreadsheet and descriptive and analytical statistics was performed with consultation of radiology and neurosurgery department statisticians using statistical product and service solution Version 15.0 software (SPSS Inc., Chicago, Illinois, USA). To simplify the comparison between diagnostic modalities, segmental arterial anatomy was reclassified into vascular distributions: anterior, middle, and posterior. Vasospasm affecting the internal carotid artery was classified as anterior and middle. Crosstabs analysis was performed with χ2/Fisher exact tests. Sensitivities and specificities were generated from 2×2 tables. Correlations were performed within each modality to determine dependence between left and right-sided data and between anterior, middle, and posterior division data. McNemar tests were performed to compare the accuracy of the tested modalities relative to one another.
Analysis of all 32 patients with angiographic vasospasm as defined in ACA, MCA, and posterior cerebral artery territories demonstrated absolute sensitivities of 96% for TCD and qualitative SPECT and 93% for semi-quantitative SPECT compared to gold standard (DSA). Evaluation of correlation between data from left and right hemispheres within each modality demonstrated high degrees of dependence, though similar evaluation between anterior, middle, and posterior divisions within each modality demonstrated non-significant dependence. Test performance characteristics were calculated (Table 1). McNemar analysis of accuracy between modalities demonstrated a better overall performance trend for qualitative SPECT versus BRASS SPECT (Fisher exact = 0.99) and a statistically significant better performance of qualitative SPECT versus TCD in measuring ACA vasospasm (Fisher exact = 0.035).
To the best of our knowledge, our study is the first to examine the utility of automated quantitative software for evaluating cerebral vasospasm on brain perfusion studies and to study the feasibility of quantification of perfusion defects on routine reporting in conjunction with qualitative analysis. We demonstrated inferior performance of BRASS analyzed SPECT relative to traditional interpretation of SPECT in overall evaluation of cerebral vasospasm. We also found that quantitative SPECT performed better than qualitative SPECT with respect to diagnosing posterior circulation vasospasm and had greater specificity in determining vasospasm affecting the anterior circulation. These results must be tempered in light of the many limitations of our study design, that is, the retrospective approach, the small sample size, and the use of only those patients with DSA proven vasospasm. The performance of TCD and both the modes of SPECT in evaluating MCA vasospasm are certainly questionable given the prevalence of DSA positive studies in that distribution. Furthermore, their respective sensitivities and specificities are below those published for TCD (sensitivity range 68–94% and specificity range 89–100%), though are within the spectrum for SPECT (sensitivity range 69–89% and specificity range 71%) [8,10,23]. This is thought, in part, to be due to our method of analyzing the segmental arterial anatomy separately and collectively. If we evaluate the modalities simply on the identification of vasospasm of any kind, sensitivities greater than 93% are generated and specificities cannot be calculated given the 100% prevalence of angiographic vasospasm. SPECT and TCD are screening tools and are expected to have sensitivities approaching unity given the clinical severity that stems from cerebral vasospasm. However, their respective ability to detect vasospasm in distinct vascular distributions varies, a point supported by our data as well as others [7,24–26].
Despite these limitations, however, meaningful conclusions can be drawn as the aforementioned pitfalls are systemic and have equally affected the evaluation of all the diagnostic modalities. The performance of BRASS SPECT relative TCD and qualitative SPECT is valid and, though inferior, still has the potential value in predicting those patients in need of intervention particularly in measuring vasospasm outside the MCA distribution. The marginally superior performance of BRASS SPECT in evaluating the anterior and posterior circulations may stem from the limited spatial resolution of SPECT and in turn qualitative visual interpretation of these areas (Fig. 1). In contrast, visual interpretation can detect interval smaller regions of hypoperfusion than the semiquantitative technique that uses counts in large regions of interest based on vascular territories in the BRASS program. Therefore, qualitative SPECT read by an experienced reader may be inherently more sensitive than BRASS, which reports the results integrated over large vascular territories. In addition, watershed defects (between territories) may also have better recognition by expert visual interpretation.
TCD and SPECT have their unique properties, which together provide a better understanding of a patient's cerebral blood flow than either in isolation. TCD is relatively inexpensive, can be performed at the bedside, does not require administration of a contrast agent or radioisotope, and has been shown to be effective in showing increased flow velocities before the clinical manifestations of ischemic vasospasm [7,8,27]. TCD is, however, limited, in that is it operator dependent, can evaluate blood flow only within the proximal portions of the intracranial arteries and thus an approximation of end perfusion, and is dependent on the acoustic window which can be absent in up to 8% of the patients [9,28,29]. TCD serves as the primary screening method for cerebral vasospasm and performs well particularly in its sensitivity for changes in MCA velocity [1,4,7]. It is, however, less sensitive for changes in the anterior and posterior circulations secondary to more technically limiting angles of insonance.
SPECT imaging, like TCD, is also noninvasive, but measures cerebral perfusion instead of flow velocities. Serial SPECT in tandem with CT can give an accurate assessment of perfusion without confounders such as hyperdynamic flow as observed in TCD. SPECT, however, is more expensive, requires patient transfer, has limited spatial resolution, and is susceptible to the issue that global decreases in blood flow may go undetected if bilateral and in all vascular territories, which may cause false negative evaluation given that qualitative interpretation often uses the contralateral or cerebellar perfusion as an internal reference . Stockbridge et al.  examined intra-observer and inter-observer variability in visual evaluation of SPECT. They found ranges of 65–100% in intra-observer performance and 29–100% in inter-observer performance, supporting the limitations of qualitative assessment . Quantitative analysis using software programs like BRASS would not be sensitive to truly balanced global changes in blood flow as global counts are used for normalization. However, there could be depiction of multiple territories of hypoperfusion if there were differences in the degree of vasospasm in the territories.
The use of quantitative software is often used in evaluating cardiac perfusion and as an adjunct to the visual interpretation to validate and/or direct further qualitative interpretation. Three-dimensional stereotactic surface projection (3D-SSP) has been used to quantify cerebral blood flow using SPECT and cerebral metabolic activity using PET in dementia patients particularly in diagnosing Alzheimer's disease [32,33]. Whereas BRASS uses a baseline individual study for comparison, 3D-SSP uses a database from normal individuals to compare with the individual scan in question. Given the heterogeneity of cerebral perfusion in the setting of SAH, whether from surgical intervention, lobar hemorrhage, or ischemic stroke, using the same individual's prior study as a baseline certainly limits these artifacts, so a truer perfusion change can be measured.
If more readily available, that is, during nights and weekends, SPECT could become a tool of the emergency department, making services such as neurology and neurosurgery more aware of its utility in patient management. Furthermore, the quantitative data generated by programs such as BRASS or 3D-SSP provide easier and potentially less biased statistical analysis. These results are better compared and/or compiled with other diagnostic modalities (e.g. CT or MR perfusion) and also to other quantifiable clinical measures (e.g. GCS, GOS, NIHSS, etc).
BRASS SPECT is a useful method for evaluating cerebral perfusion and needs further optimization particularly as it pertains to establishing meaningful semi-quantitative parameters in evaluating cerebral perfusion, which a larger, prospective study could better evaluate. It can serve as a useful adjunct to traditional SPECT evaluation of SAH particularly in determining subtle changes in the perfusion of the anterior and posterior cerebral artery distributions.
1. Komotar R, Zacharia B, Valhora R, Mocco J, Connolly EJ. Advances in vasospasm treatment and prevention. J Neurol Sci 2007; 261:134–142.
2. Brisman J, Eskridge J, Newell D. Neurointerventional treatment of vasospasm. Neurol Res 2006; 28:769–776.
3. Broderick J, Brott T, Tomsick T, Tew J, Duldner J, Huster G. Management of intracerebral hemorrhage in a large metropolitan population. Neurosurgery 1994; 34:882–887. discussion 887.
4. Zwienenberg-Lee M, Hartman J, Rudisill N, Muizelaar J. Endovascular management of cerebral vasospasm. Neurosurgery 2006; 59 (Suppl 3):S139–S147. discussion S3–S13.
5. Komotar R, Zacharia B, Otten M, Mocco J, Lavine S. Controversies in the endovascular management of cerebral vasospasm after intracranial aneurysm rupture and future directions for therapeutic approaches. Neurosurgery 2008; 62:897–905. discussion 905–907.
6. Gonzalez N, Boscardin W, Glenn T, Vinuela F, Martin N. Vasospasm probability index: a combination of transcranial Doppler velocities, cerebral blood flow, and clinical risk factors to predict cerebral vasospasm after aneurysmal subarachnoid hemorrhage. J Neurosurg 2007; 107:1101–1112.
7. Kincaid M, Souter M, Treggiari M, Yanez N, Moore A, Lam A. Accuracy of transcranial Doppler ultrasonography and single-photon emission computed tomography in the diagnosis of angiographically demonstrated cerebral vasospasm. J Neurosurg 2009; 110:67–72.
8. Egge A, Sjøholm H, Waterloo K, Solberg T, Ingebrigtsen T, Romner B. Serial single-photon emission computed tomographic and transcranial Doppler measurements for evaluation of vasospasm after aneurysmal subarachnoid hemorrhage. Neurosurgery 2005; 57:237–242. discussion 237–242.
9. Lewis D, Newell D, Winn H. Delayed ischemia due to cerebral vasospasm occult to transcranial Doppler. An important role for cerebral perfusion SPECT. Clin Nucl Med 1997; 22:238–240.
10. Rajendran J, Lewis D, Newell D, Winn H. Brain SPECT used to evaluate vasospasm after subarachnoid hemorrhage: correlation with angiography and transcranial Doppler. Clin Nucl Med 2001; 26:125–130.
11. Powsner R, O'Tuama L, Jabre A, Melhem E. SPECT imaging in cerebral vasospasm following subarachnoid hemorrhage. J Nucl Med 1998; 39:765–769.
12. Suarez J, Qureshi A, Yahia A, Parekh, PD, Tamargo, RJ, Williams, MA, et al. Symptomatic vasospasm diagnosis after subarachnoid hemorrhage: evaluation of transcranial Doppler ultrasound and cerebral angiography as related to compromised vascular distribution. Crit Care Med 2002; 30:1348–1355.
13. Becker G, Greiner K, Kaune B, Winkler J, Brawanski A, Warmuth-Metz M, Bogdahn U. Diagnosis and monitoring of subarachnoid hemorrhage by transcranial color-coded real-time sonography. Neurosurgery 1991; 28:814–820.
14. Radau P, Linke R, Slomka P, Tatsch K. Optimization of automated quantification of 123I-IBZM uptake in the striatum applied to parkinsonism. J Nucl Med 2000; 41:220–227.
15. Radau P, Slomka P, Julin P, Svensson L, Wahlund L. Evaluation of linear registration algorithms for brain SPECT and the errors due to hypoperfusion lesions. Med Phys 2001; 28:1660–1668.
16. Slomka P, Radau P, Hurwitz G, Dey D. Automated three-dimensional quantification of myocardial perfusion and brain SPECT. Comput Med Imaging Graph 25:153–164.
17. Bosman T, Van Laere K, Santens P. Anatomically standardized 99mTc-ECD brain perfusion SPET allows accurate differentiation between healthy volunteers, multiple system atrophy and idiopathic Parkinson's disease. Eur J Nucl Med Mol Imaging 2003; 30:16–24.
18. Van Laere KJ, Warwick J, Versijpt J, Goethals I, Audenaert K, Van Heerden B, Dierckx R. Analysis of clinical brain SPECT data based on anatomic standardization and reference to normal data: an ROC-based comparison of visual, semiquantitative, and voxel-based methods. J Nucl Med 2002; 43:458–469.
19. Lobotesis K, Fenwick J, Phipps A, Ryman A, Swann A, Ballard C, et al. Occipital hypoperfusion on SPECT in dementia with Lewy bodies but not AD. Neurology 2001; 56:643–649.
20. Hurwitz GA, Hurwitz RC, Slomka PJ, Radau PE. Demarcation of vascular lesions on heart and brain SPECT: fabric art representation. J Nucl Cardiol 2002; 9:572.
21. Bradley K, O'Sullivan V, Soper N, Nagy Z, King EM, Smith AD, Shepstone BJ. Cerebral perfusion SPET correlated with Braak pathological stage in Alzheimer's disease. Brain 2002; 125 (Pt 8):1772–1781.
22. Deutsch G, Mountz J, Katholi C, Liu H, Harrell L. Regional stability of cerebral blood flow measured by repeated technetium-99m-HMPAO SPECT: implications for the study of state-dependent change. J Nucl Med 1997; 38:6–13.
23. Rigamonti A, Ackery A, Baker A. Transcranial Doppler monitoring in subarachnoid hemorrhage: a critical tool in critical care. Can J Anaesth 2008; 55:112–123.
24. Sviri G, Mesiwala A, Lewis D, Britz GW, Nemecek A, Newell DW, et al. Dynamic perfusion computerized tomography in cerebral vasospasm following aneurysmal subarachnoid hemorrhage: a comparison with technetium-99m-labeled ethyl cysteinate dimer-single-photon emission computerized tomography. J Neurosurg 2006; 104:404–410.
25. Sviri G, Ghodke B, Britz G, Douville CM, Haynor DR, Mesiwala AH, et al. Transcranial Doppler grading criteria for basilar artery vasospasm. Neurosurgery 2006; 59:360–366. discussion 360–366.
26. Newell D, Grady M, Eskridge J, Winn H. Distribution of angiographic vasospasm after subarachnoid hemorrhage: implications for diagnosis by transcranial Doppler ultrasonography. Neurosurgery 1990; 27:574–577.
27. Aaslid R, Markwalder T, Nornes H. Noninvasive transcranial Doppler ultrasound recording of flow velocity in basal cerebral arteries. J Neurosurg 1982; 57:769–774.
28. Sviri GE, Britz GW, Lewis DH, Ghodke B, Mesiwala AH, Haynor DH, Newell DW. Brainstem hypoperfusion in severe symptomatic vasospasm following aneurysmal subarachnoid hemorrhage: role of basilar artery vasospasm. Acta Neurochir (Wien) 2006; 148:929–934. discussion 934–925.
29. Sviri G, Newell D, Lewis D, Douville C, Ghodke B, Chowdhary M, et al. Impact of basilar artery vasospasm on outcome in patients with severe cerebral vasospasm after aneurysmal subarachnoid hemorrhage. Stroke 2006; 37:2738–2743.
30. Rosen J, Butala A, Oropello J, Sacher M, Rudolph SH, G oldsmith SJ, et al. Postoperative changes on brain SPECT imaging after aneurysmal subarachnoid hemorrhage. A potential pitfall in the evaluation of vasospasm. Clin Nucl Med 1994; 19:595–597.
31. Stockbridge H, Lewis D, Eisenberg B, Lee M, Schacher S, van Belle G, et al. Brain SPECT: a controlled, blinded assessment of intra-reader and inter-reader agreement. Nucl Med Commun 2002; 23:537–544.
32. Minoshima S, Frey K, Koeppe R, Foster N, Kuhl D. A diagnostic approach in Alzheimer's disease using three-dimensional stereotactic surface projections of fluorine-18-FDG PET. J Nucl Med 1995; 36:1238–1248.
33. Tang B, Minoshima S, George J, Robert A, Swine C, Laloux P, Borght TV. Diagnosis of suspected Alzheimer's disease is improved by automated analysis of regional cerebral blood flow. Eur J Nucl Med Mol Imaging 2004; 31:1487–1494.