Assessment of Thrombus in Acute Middle Cerebral Artery Occlusion Using Thin-Slice Nonenhanced Computed Tomography Reconstructions
Background and Purpose— We sought to evaluate how accurately length and volume of thrombotic clots occluding cerebral arteries of patients with acute ischemic stroke can be assessed from nonenhanced CT (NECT) scans reconstructed with different slice widths.
Methods— NECT image data of 58 patients with acute ischemic stroke with vascular occlusion proven by CT angiography were reconstructed with slice widths of 1.25 mm, 2.5 mm, 3.75 mm, and 5 mm. Thrombus lengths and volumes were quantified based on these NECT images by detecting and segmenting intra-arterial hyperdensities. The results were compared with reference values of thrombus length and volume obtained from CT angiography images using Bland-Altman analysis and predefined levels or tolerance to find NECT slice thicknesses that allow for sufficiently accurate thrombus quantification.
Results— Thrombus length can be measured with high accuracy using the hyperdense middle cerebral artery sign detected in NECT images with slice thicknesses of 1.25 mm and 2.5 mm. We found mean deviations from the reference values and limits of agreement of −0.1 mm±0.6 mm with slice widths of 1.25 mm and 0.1 mm±0.7 mm for slice widths of 2.5 mm. Thrombus length measurements in NECT images with higher slice width and all evaluated thrombus volume measurements exhibited severe dependence on the level and did not match the accuracy criteria.
Conclusion— The length of the hyperdense middle cerebral artery sign as detected on thin-slice NECT reconstructions in patients with acute ischemic stroke can be used to quantify thrombotic burden accurately. Thus, it might qualify as a new diagnostic parameter in acute stroke management that indicates and quantifies the extent of vascular obliteration.
- acute care
- acute stroke
- embolic stroke
- stroke care
- stroke management
In acute anterior ischemic stroke, nonenhanced CT (NECT) may demonstrate a hyperdense middle cerebral artery sign (HMCAS) as a highly specific marker of thrombotic vascular occlusion. So far, the sign has not been used to assess clot burden because standard NECT slices are typically too thick to accurately delineate thrombus. Using thin-slice reconstructions of standard NECT data, thrombotic clots should contrast better with surrounding tissue due to diminished volume averaging effects. The high spatial resolution of such reconstructions can even outweigh their low signal-to-noise levels and therefore permit quantifying intravascular thrombus by accurate segmentation of the HMCAS.
The HMCAS was discovered >25 years ago.1 Since then, authors have pointed out that the sign is highly specific for thrombus occluding cerebral arteries2 and that it predicts poor outcome of intravenous thrombolysis in acute strokes.3–5 Outcome furthermore depends on the location of the intravascular hyperdensity and on its size.6 The smaller, peripherally located variant of the HMCAS, the middle cerebral artery dot sign, is associated with a better outcome of intravenous thrombolysis compared with the HMCAS.7,8 However, a recent study reports that patients with HMCAS with acute stroke benefited more from intra-arterial than from intravenous thrombolysis.9 Thus, future decisions about the optimal therapy for acute strokes should take into consideration the total extent of vascular occlusion by thrombus. Recently, a technique for measuring the volume of intravascular thrombus using thin-section NECT scans was developed.10 Using additional high-resolution CT scans, the authors were able to detect thrombus in nearly all cases of acute stroke in which CT angiography had proven occlusions of the middle cerebral artery an internal carotid artery. Unfortunately, this quantitative method was not validated by any references such as CT angiography or MRI. Furthermore, the technique required additional scanning and hence extra radiation exposure and time.
To make this new tool relevant for imaging in acute ischemic stroke, we have to comply with 2 requirements. First, thrombus should be depicted from standard NECT scan data to avoid additional scanning. On multidetector row CT scanners, this can be realized with thin-section reconstruction of NECT data scanned using standard protocols. Second, we have to determine how precisely thrombus extent can be delineated using these images with high spatial resolution but low signal-to-noise ratio. To determine the true extent of vascular obliteration, CT angiography images with intravascular contrast proximal and distal to the obstructed site can be used as a reference. The purpose of this study was to determine how accurately thrombus volume and length can be measured using thin CT slice reconstructions of standard NECT scans.
Between April 2008 and March 2009, we monitored a group of patients who presented with acute stroke in the middle cerebral artery (MCA) territory within 6 hours from symptom onset in a prospective case series. Patients were only included when pretreatment NECT was followed by CT angiography (CTA), which had to prove occlusion of the most proximal segment (M1) of the MCA. Furthermore, 2 neuroradiologists reading the NECT and CTA images independently had to agree on a close match between the location and spatial extent of focal hyperdensities representing thrombus in NECT images and intravascular contrast voids observed in the corresponding CTA images (Figure 1A–B). This means that in all included patients, sufficient retrograde blood flow across collateral vessels allowed for contrasting the obliterated vessel distally to the site of occlusion.
Demographic data of all included patients were recorded.
A multidetector row CT scanner with 64 detector rows (Brilliance 64; Philips, Best, The Netherlands) was used for NECT and CTA scanning in all patients. The standard NECT protocol entailed a collimation of 16×0.625 mm, a tube voltage of 120 kV, a tube current of 320 mAs, and selection of a high-resolution focal spot. Incremental scanning and a smooth reconstruction kernel, optimized for brain imaging, yielded 2.5-mm thick NECT slices. Cranial CTA was initiated by visual contrast bolus tracking in the cervical arteries after infusion of an 80-mL bolus of 350 mg I/mL at a rate of 5 mL/s followed by 40 mL of saline flush. The vessels were scanned in helical mode with a volume pitch of 1.2 and a collimation of 64×0.625 mm, a tube voltage of 80 kV, and a tube current of 280 mAs.
All NECT images were reconstructed offline with a slice width of 0.625 mm. All other reconstruction parameters of the standard cranial NECT protocol were kept constant. Subsequently, these thin slices were resampled using a B-Spline interpolator11 for gantry tilt correction. After resampling, additional reconstructions with slice widths of 1.25 mm, 2.5 mm, 3.75 mm, and 5 mm were calculated using averages of the original thin slices.
Thrombus Segmentation and Quantitative Analysis of Thrombotic Burden in NECT Images
Thrombus size was measured after semiautomated segmentation. Therefore, 1 neuroradiologist manually had to define regions of interest around the course of cerebral arteries of the anterior circulation in all preprocessed CT data sets. In these regions, every image voxel with a density between 55 and 80 Hounsfield units was regarded as a potential seed for a subsequent segmentation using a seeded region-growing algorithm. This segmentation process included all seed neighbor pixels with a density between 45 and 80 Hounsfield units. These conditions allowed for maximizing segmented thrombus volume without leakage affecting the surrounding tissue.10 Thrombus volume was computed using the segmented object by multiplying the number of object voxels by the voxel dimensions. To derive the thrombus length, the segmentation result first had to be reduced to a medial axis representation. Therefore, it was reduced to a skeleton by applying a topology-preserving morphological thinning operation. Finally, the maximum euclidean length of the resulting skeleton was calculated. The erosion distance of the skeletonization operation was added twice to this length to account for the shortening of the skeleton with respect to the thrombus by the morphological thinning procedure. Examples of the results of the segmentation and skeletonization algorithm are shown in Figure 1C.
To define a reference value for thrombotic burden, the occluded arterial segment was measured using the CTA images along with the NECT images. For this purpose, the CTA and NECT images first had to be registered using a rigid 3-dimensional affine transform. Next, a medial axis of the occluded arterial segment was manually defined. Therefore, a polygon was drawn by connecting points in the vessel center proximally and distally to the occlusion site. Because the occluded segments could not be expected to follow a straight line between the patent proximal and distal vascular segments, their best medial axis representations had to be defined by using the hyperdense artery signs that were superimposed onto the CTA images by the preceding registration.
A B-Spline approximation of the polygon was used to smooth the resulting vascular axis. This medial axis was examined from 3 orthogonal maximum intensity projection images and eventually corrected to align accurately with the vascular axis. Finally, this skeleton representation of the artery was cut at the proximal and distal ends of the occlusion site defined by the void of intravascular contrast. The resulting segment length was used as a reference length of the thrombus. To define a reference volume for each thrombus, the HMCAS was quantified using NECT images reconstructed with a slice thickness of 0.625 mm. Thrombus volume was quantified in the same way as in NECT images with slice widths between 1.25 mm and 5 mm as described previously. Two neuroradiologists independently measured the reference lengths and volumes of all clots to evaluate interobserver variability.
Data analysis started with processing the patients’ personal statistics. In the next step, we defined tolerance levels for the measurements of thrombus length and thrombus volume. Based on a typical NECT voxel width of approximately 0.5 mm in the axial plane, we considered a tolerance level of ±1 mm appropriate because this corresponds to an error of 1 voxel on each side of the clot. Assuming a typical diameter of the MCA of 2.5 mm and a tolerance of ±1 mm for thrombus length measurements, we defined a tolerance level for clot volume measurements of ±10 mm3. Subsequently, the Bland and Altman method was used to analyze the interobserver errors of the reference measurement results of thrombus volume and length. Therefore, the differences between the measurements by both observers were used to calculate the mean of the differences and upper and lower values of the 95% limits of agreement. The same technique was used to compare thrombus lengths and volumes in groups of equal CT slice width with the associated reference values. For every NECT slice width, the spread of these differences was evaluated with Bland-Altman plots for independence of the magnitude of thrombus volume and thrombus length. In cases of independence, the CO for the upper and lower limits of agreement was calculated using a pairwise t test. Finally, the mean of the differences and the upper and lower values of the 95% limits of agreement were compared with the tolerance levels defined above to find those NECT slice widths that allow for measuring thrombus length and volume with sufficient precision compared with the reference technique.
A total of 58 people matched the inclusion criteria. Thirty-six patients were male with a median age of 62 and an age range from 43 to 79 years. In all included cases, CTA imaging proved MCA obliteration. All patients had either pure obliteration of the MCA main stem or from occlusion of the MCA main stem with thrombus reaching the supraclinoid segment of the ipsilateral internal carotid artery.
In all cases, preprocessing of the NECT image data by selecting a region of interest around the course of the anterior circulation vessels resulted in compact seeds within the intravascular hyperdensities. Seeded region growing resulted in well-defined thrombus representations in all patients with NECT slice widths of 1.25 mm and 2.5 mm. When a NECT slice thickness of 3.75 mm was chosen, no thrombus was detected in 4 of the 58 patients (6.9%). With a NECT slice thickness of 5 mm, clots were missed in 14 patients (24.1%).
Quantitative Thrombus Analysis
The interobserver variability for the reference data of thrombus length and volume was found to be within the limits of the predefined tolerance levels. The mean thrombus length deviation was 0.2 mm (limits of agreement: ±0.6 mm); the mean volume deviation was 2.3 mm3 (limits of agreement: ±4.3 mm3).
The magnitude-dependent differences between thrombus lengths measured in different NECT reconstructions on the 1 hand and thrombus length reference data on the other are demonstrated in Bland-Altman plots in Figure 2. The equivalent comparisons for thrombus volume measurements are shown in Figure 3. For all measurements of thrombus volumes and lengths, the mean deviations and the limits of agreement as well as the CIs of the limits of agreement are listed in the Table.
According to Figure 2, thrombus length measurements are only independent of the level when they are based on NECT images reconstructed with a slice thickness of 1.25 mm or 2.5 mm. Only under these conditions, the limits of agreement and their upper and lower confidence levels as estimated by the t test are within the limits of the predefined tolerance level of ±1 mm (see the Table). Furthermore, the mean deviation between NECT measurements and reference lengths is very close to zero. As the NECT slice thickness increases beyond 2.5 mm, the mean deviation of thrombus length rises above the upper limit of the tolerance levels and the level of agreement exceeds the tolerance level by more than a factor of 4.
The measurements of thrombus volumes from NECT images with slice widths between 1.25 mm and 5 mm all deviate from the reference data with mean differences exceeding the defined tolerance levels. Thus, none of the evaluated NECT data sets can be used to evaluate thrombus volume with sufficient precision compared with the reference data.
According to our results, reconstructions of standard cranial NECT data with a slice width of ≤2.5 mm using simple image postprocessing tools allow for the accurate measurement of the length of a thrombus in a cerebral artery of the circle of Willis. However, the same tools fail to identify the volume of thrombus with sufficient precision with all other NECT slice widths we evaluated. The latter finding is most likely due to the different influences of partial volume effects on the spatial extent of the HMCAS as slice thickness increases. However, partial volume averaging only has a minor impact on the middle axis representation of the thrombus used for length measurements. It only degrades the middle axis significantly as the CT slice width increases beyond 2.5 mm.
Thus, thrombotic burden can be described accurately by thrombus length if standard cranial NECT scanning protocols with multidetector row CT scanners are used. It requires only a few seconds of extra reconstruction time and no additional scanning, which ultimately saves radiation dose and time needed for subsequent treatment. Furthermore, because a slice width of 2.5 mm still permits sufficiently accurate delineation of thrombus extent, this slice width could initially be used as a reconstruction parameter for the standard NECT protocol.
A recently published technique for measuring thrombus volume for the first time from thin-slice NECT image quantitatively defined thrombotic burden in acute strokes.10 These results are highly encouraging for establishing a valuable new prognostic parameter in acute strokes. This might even eventually guide therapeutic decisions. Our study results show that the same technique is applicable to standard thin-slice NECT data obviating additional scanning. We furthermore confirm the gray scale thresholds described for thrombus segmentation by seeded region-growing. These correspond well with density values of thrombus reported in earlier studies.12 In contrast to the preceding work on volumetric assessment of thrombus, our results show that thrombotic burden might be more accurately defined by thrombus length than by volume. Thrombus extent along the vascular axis very recently has been used as a prognostic parameter for patient outcome using scoring systems for occlusion on CTA images.13,14 These systems associate scores with different vessel segments. These scoring systems have been shown to correlate well with treatment success and patient outcome. Our technique has the potential to advance these systems to describe thrombotic burden in a more reproducible way.
Thus, we have defined a new parameter for the management of acute ischemic stroke that directly describes the cause of patients’ states and symptoms. The technique can be applied immediately at most stroke units using modern multidetector row CT scanners with high-resolution detectors. The required thin-slice NECT images can simply be reconstructed using standard scanning protocols. Therefore, the data required for thrombus length analysis will not necessitate additional radiation and the time needed for 1 additional reconstruction is negligible. Because current studies aim to define simple clinical and NECT imaging parameters to guide patients in acute stroke,15 quantification of thrombotic burden might become included as a standard parameter into the guidelines for management of acute ischemic stroke.16 Thrombus length might particularly be useful as a parameter to decide in which patients intravenous thrombolysis will most likely fail to recanalize the occluded vessels.17 If this parameter is generally accepted, it might even be used to evaluate the success of therapy by detecting and measuring residual thrombus.18,19
A major potential pitfall when applying the thrombus segmentation technique might result from conditions different from thrombotic vascular occlusion leading to hyperdensities of intracranial vessels. Calcified arterial walls and a high hematocrit could cause vascular hyperdensities resembling thrombus. This underlines the limitations of our study. In the relatively small patient population we studied, thrombotic vascular occlusion was well defined by arterial hyperdensities in all cases. In none of the patients was segmentation of intraluminal thrombus due to arterial calcification or an increase in x-ray attenuation in the bloodstream a problem. Because both of these conditions have to be expected in a larger patient population, it is highly important to apply the technique in a larger prospective study. Furthermore, it would be very helpful if a reference standard for thrombus length measurements could be established that would be completely independent from the NECT and CTA imaging. This reference might use MR sequences optimized for displaying thrombotic clots.
In summary, standard NECT protocols for imaging of acute stroke can easily be extended by thin-slice reconstructions that allow for assessing clot burden accurately by thrombus length. The occluded length can be used to predict patient outcome and the success rate of intravenous lysis. If subsequent studies define a limit of thrombus length above which intravenous thrombolysis most likely fails because the lytics simply never successfully penetrate the whole thrombus extent, then this limit will define a threshold above which intra-arterial thrombolysis would have to be strongly considered. Thus, further studies will have to show the feasibility of the technique we described in a larger patient population. This population should be analyzed further with respect to their treatment plans and clinical outcomes to decide whether intra-arterial hyperdensities can be used to define the best treatment option in acute ischemic stroke.
- Received January 31, 2010.
- Revision received March 12, 2010.
- Accepted March 22, 2010.
Gács G, Fox AJ, Barnett HJ, Vinuela F. CT visualization of intracranial arterial thromboembolism. Stroke. 1983; 14: 756–762.
von Kummer R, Meyding-Lamade U, Forsting M, Rosin L, Rieke K, Hacke W, Sartor K. Sensitivity and prognostic value of early CT in occlusion of the middle cerebral artery trunk. AJNR Am J Neuroradiol. 1994; 15: 9–15.
Leys D, Pruvo JP, Godefroy O, Rondepierre P, Leclerc X. Prevalence and significance of hyperdense middle cerebral artery in acute stroke. Stroke. 1992; 23: 317–324.
Moulin T, Cattin F, Crépin-Leblond T, Tatu L, Chavot D, Piotin M, Viel JF, Rumbach L, Bonneville JF. Early CT signs in acute middle cerebral artery infarction: predictive value for subsequent infarct locations and outcome. Neurology. 1996; 47: 366–375.
Tomsick T, Brott T, Barsan W, Broderick J, Haley EC, Spilker J, Khoury J. Prognostic value of the hyperdense middle cerebral artery sign and stroke scale score before ultraearly thrombolytic therapy. AJNR Am J Neuroradiol. 1996; 17: 79–85.
Leary M. Validation of computed tomographic middle cerebral artery ‘dot’ sign: an angiographic correlation study. Stroke. 2003; 34: 2636–2640.
Barber PA, Demchuk AM, Hudon ME, Pexman JH, Hill MD, Buchan AM. Hyperdense sylvian fissure MCA ‘dot’ sign: a CT marker of acute ischemia. Stroke. 2001; 32: 84–88.
Mattle H, Arnold M, Georgiadis D, Baumann C, Nedeltchev K, Benninger D, Remonda L, Von Budingen C, Diana A, Pangalu A, Schroth G, Baumgartner R. Comparison of intraarterial and intravenous thrombolysis for ischemic stroke with hyperdense middle cerebral artery sign. Stroke. 2008; 39: 379–383.
Kim E, Yoo E, Choi H, Lee J, Heo J. Thrombus volume comparison between patients with and without hyperattenuated artery sign on CT. AJNR Am J Neuroradiol. 2008; 29: 359–362.
Barreto AD, Albright KC, Hallevi H, Grotta JC, Noser EA, Khaja AM, Shaltoni HM, Gonzales NR, Illoh K, Martin-Schild S, Campbell MS, Weir RU, Savitz SI. Thrombus burden is associated with clinical outcome after intra-arterial therapy for acute ischemic stroke. Stroke. 2008; 39: 3231–3235.
Kobayashi A, Wardlaw JM, Lindley RI, Lewis SC, Sandercock PAG, Czlonkowska A; IST-3 Collaborative Group. Oxfordshire Community Stroke Project clinical stroke syndrome and appearances of tissue and vascular lesions on pretreatment CT in hyperacute ischemic stroke among the first 510 patients in the Third International Stroke Trial (IST-3). Stroke. 2009; 40: 743–748.
Adams H, Del Zoppo G, Alberts M, Bhatt D, Brass L, Furlan A, Grubb R, Higashida R, Jauch E, Kidwell C, Lyden P, Morgenstern L, Qureshi A, Rosenwasser R, Scott P, Wijdicks E. Guidelines for the early management of adults with ischemic stroke: a guideline from the American Heart Association/American Stroke Association Stroke Council, Clinical Cardiology Council, Cardiovascular Radiology and Intervention Council, and the Atherosclerotic Peripheral Vascular Disease and Quality of Care Outcomes in Research Interdisciplinary Working Groups: the American Academy of Neurology affirms the value of this guideline as an educational tool for neurologists. Circulation. 2007; 115: e478–e534.
Kharitonova T, Ahmed N, Thorén M, Wardlaw JM, von Kummer R, Glahn J, Wahlgren N; SITS Investigators. Hyperdense middle cerebral artery sign on admission CT scan—prognostic significance for ischaemic stroke patients treated with intravenous thrombolysis in the safe implementation of thrombolysis in Stroke International Stroke Thrombolysis Register. Cerebrovasc Dis. 2009; 27: 51–59.
Kharitonova T, Thorén M, Ahmed N, Wardlaw J, von Kummer R, Thomassen L, Wahlgren N; SITS Investigators. Disappearing hyperdense middle cerebral artery sign in ischaemic stroke patients treated with intravenous thrombolysis: clinical course and prognostic significance. J Neurol Neurosurg Psychiatry. 2009; 80: 273–278.
Saqqur M, Molina CA, Salam A, Siddiqui M, Ribo M, Uchino K, Calleja S, Garami Z, Khan K, Akhtar N, O'Rourke F, Shuaib A, Demchuk AM, Alexandrov AV; CLOTBUST Investigators. Clinical deterioration after intravenous recombinant tissue plasminogen activator treatment: a multicenter transcranial Doppler study. Stroke. 2007; 38: 69–74.