Neuroradiology (1996) 38:245 251 9 Springer-Verlag 1996
J. Rieger N.Hosten K. Neumann R. Langer P. Molscn W. R. Lanksch K. J. Pfeifer R. Felix
Received: 13 April 1994 Accepted: 28 February 1995 The authors would like to dedicate this paper to Prof. Dr. sc. reed. R. Lehmann
J. Rieger ( ~ ) 9K. J. Pfeifer Abteilung fiir Radiologie, Chirurgische Universitfitsklinik des Klinikums Innenstadt, Nussbaumstrasse 20, D-80336 Munich, Germany N. Hosten 9K. Neumann 9R. Langer E Molsen 9R. Felix Abteilung fiir Radiologie, Rudolf Virchow Universitfitsklinik, Freie Universitfit Berlin W. R. Lanksch Abteilung fiir Neurochirurgie, Rudolf Virchow Universitfitsklinik, Freie Universit~it Berlin
Initial clinical experience with spiral CT and 3D arterial reconstruction in intracranial aneurysms and arteriovenous malformations
Abstract We studied 32 consecutive patients with known or suspected cerebrovascular abnormalities studied with spiral CT following a intravenous bolus injection of iodinated contrast medium with a power injector. Flow was 3 or 4 ml/s. In an attempt to define the appropriate delay time and scan duration a cranial angio-CTwithout table increment was performed on 10 patients. E n h a n c e m e n t was measured by manually placed regions of interest within the left middle cerebral artery and the inferior sagittal sinus. All patients except one had intraarterial angiography (DSA) for comparison. In 6 patients with an arteriovenous malformation ( A V M ) follow-up was possible after one and/or two embolisation procedures. These patients had plain and contrast-enhanced spiral CT. The diagnosis was aneurysm in 9 (8 berry aneurysms, one giant fusiform aneurysm), A V M in 13 (all supratentorial) and traumatic arteriovenous fistula in one. In 9 patients there
Introduction The new technique of spiral computed tomography (SCT) combines a continuously rotating tube with continuous patient transport through the gantry at a determined table speed. Depending on the freely selectable table speed SCT allows fast, gapless data acquisition of variable body volumes. Respiration - induced misregistration of pathological findings such as small pulmonary
were no detectable pathological vascular findings. After 3D reconstruction the size (between 5 and 28 mm), location and the relationship to the parent vessel of the aneurysms, the extent of the AVMs and the distribution of the embolisation material could be demonstrated clearly. The main feeding vessel(s), nidus and draining veins were reliably shown. The decreased extent of the AVMs after embolisation was clearly demonstrated. There was no difference in diagnosis when D S A and 3 D - C T w e r e compared by two independent radiologists. We consider arterial spiral CT with 3D reconstruction to have the potential of offering important diagnostic information for the treatment of intracranial AVMs and aneurysms.
Key words Spiral computed t o m o g r a p h y . Three-dimensional reconstruction. Intracranial aneurysms. Arteriovenous malformations
nodules can be avoided by the possibility of scanning during a single breath-hold. The quality of 3D reconstructions is not compromised by patient motion or inconsistent inspiration levels as in discontinuous scanning and can be increased with decreasing table speed, thus gaining an increasing amount of data per volume measured. The arterias of the circle of Willis invite this new technique because at a minimal possible table speed of 1 mm/s one or two 24 s SCT acquisitions nor-
246
m a l l y c o v e r the w h o l e r e g i o n of interest, in which m a n y a n e u r y s m s lie a n d a m a x i m u m of f o u r a c q u i s i t i o n s are n e c e s s a r y to c o m p l e t e l y cover e v e n large a r t e r i o v e n o u s m a l f o r m a t i o n s ( A V M s ) . P a t i e n t s with s u s p e c t e d intrac r a n i a l b l e e d i n g who u n d e r g o c r a n i a l C T do n o t n e e d to be t r a n s p o r t e d a n d c o n t r a s t - e n h a n c e d S C T takes little a d d i t i o n a l time. I n c o m b i n a t i o n with a p o w e r injector, the S C T t e c h n i q u e m a k e s it possible to follow a b o l u s of c o n t r a s t m e d i u m within the short p e r i o d of m a x i m u m intravascular concentration. Threshold-based postprocessing of the raw S C T d a t a can b e p e r f o r m e d o n a b a c k g r o u n d w o r k s t a t i o n w i t h o u t i n t e r f e r i n g with a n y following C T studies. T h e final 3D i m a g e s are d i s p l a y e d with the s h a d e d surface display (SSD) t e c h n i q u e .
Table 1 Technical factors
Mode Algorithm kV mA/s Scan time Table speed Slice thickness Increment Delay time Contrast medium
Spiral CT Soft 120 165 24 s 1 mm/s i mm 1 mm 14-17 s 60-80 ml each bolus injection
Diagram arterial
250 200
Patients and methods .-~
150
Initial studies 100
The orientating definition of the optimum scanning parameters was performed on 10 patients aged between 39 and 73 years with intracerebral metastatic disease, without regard to possible differences in circulation times. On a third-generation CT scanner, unenhanced CT images were obtained. We selected a representative image of the skull base which showed both the middle cerebral artery (MCA) and the inferior sagittal sinus and performed a 2 mm angio-CI'with no table increment. This provided nine scans with a fixed interscan time of 2 s. The delay time was 10 s once bolus injection of 80 ml iodinated contrast medium was started with a power injector, at 4 ml/s. A further four scans with a interscan time of 9 s followed. Density measurements were made in manually placed regions of interest within the left MCA and the inferior sagittal sinus for each scan and a time/attenuation value diagram was constructed.
Patients There were 26 patients referred from the neurosurgical or emergency departments and 6 from municipal hospitals. The intra-arterial angiography (DSA) and embolisation procedures were performed using digital vascular imaging; histoacrylate combined with tantalum was used as embolisation material. To find the optimum level for starting SCT, plain scans of the skull base, 5 mm thick were obtained. Depending on the size of the aneurysm or AVM 1-4 SCTwere obtained consecutively. With our CT apparatus a maximum of 24 360 ~ tube rotations of i s is possible for data acquisition in the spiral mode. Table feed is freely selectable between 1 and 10 ram, allowing a maximum distance of 22 mm to 22 cm to be covered during a single acquisition (Table 1). The implemented software for the 3D reconstruction of the SCT data obtained is based on a thresholding technique. This requires that the user defines a specific range of attenuation values according to the attenuation values of the tissue of interest. In this binary system any voxel with an attenuation value outside the range selected is regarded as not representing this tissue; any given voxel is taken to contain either 0 % or 100 % of a given tissue, leading to one of the most important limitations of this imagerendering technique: voxels with different tissues, e. g., muscle and bone, cause averaging of attenuation values and cannot be correctly classified.
soi 0 10
15
20
25
30
35
40
45
50
55
time in s e c o n d s p. i.
Fig.1 Time-attenuation curves (p.i. postinjection) Before starting the second or third SCT the single level images must be calculated from the raw SCT data; otherwise the whole data set would be lost. This proved not to be a disadvantage, because the calculation time was needed for washout of the contrast medium. After a minimum of 2 min the next SCT acquisition was started, with one overlapping slice. The next step requires selection of the appropriate threshold values to confine the image to only the desired structures. We chose a single threshold value of 120 + 10 Hounsfield units (HU) which clearly exceeded the attenuation values of normal brain parenchyma or a haemorrhage. Definition of an upper threshold was not necessary, because the bone of the skull base should be visible for orientation. The reconstruction matrix had 512 x 512 pixels. Every voxel on every single image is evaluated. Obscuring anatomical structures below the predefined threshold are thereby eliminated on every single image before creating the 3D reconstruction with SSD. Postprocessing options for improved demonstration of pathological findings include a cut function, a simulated light source and the possibility of freely rotating the 3D image in any direction. With these tools a minimum of four different projections was created for assessment of every patient. The raw SCT data of two patients were additionally filed to a transputer which consists of several pipeline processors with a calculation capacity of 800 million instructions/s. Postprocessing possibilities are thereby widened by the facility (among others) of colour-coding tissues with different attenuation values.
Results F i g u r e 1 shows the t i m e - r e l a t e d a t t e n u a t i o n v a l u e s w i t h i n the left M C A a n d i n f e r i o r sagittal sinus after bolus i n j e c t i o n of c o n t r a s t m e d i u m . T h e s e d a t a led us to a n
247
shown
Fig.3a-r "Butterfly"-shaped giant fusiform aneurysm of the right middle cerebral artery (MCA) in a 58-year-old patient. a SSD image: the aneurysm (arrow) and the main branching artery are well delineated. The major thrombosed part of the aneurysm is not seen but could be identified on the standard CT images, b Intra-arterial DSA, right anterior oblique projection. The higher resolution enables demonstration of even small arterial branches. The real size of the aneurysm is not shown c The postprocessing possibilities are demonstrated on this SSD image: having rotated the image and "cut away" part of the skull better demonstration of the aneurysm (arrow) is achieved
248
Fig.4 a An aneurysm (arrow) of the right MCA trifurcation is shown on this SSD image. The branching arteries are clearly delineated, b The aneurysm (arrow) is shown on intra-arterial DSA, right anterior projection
appropriate delay time of 14-17 s. In the spiral mode the scanner offers the possibility of determining a scanning time between 12 and 24 s, providing 12-24 360 ~ tube rotations. Although the attenuation values of the larger intracranial veins exceed the selected threshold during a 24-s SCT acquisition we chose the maximum scanning time because the main part of these veins is located close to the skull. Identification of these veins is therefore easy and they do not interfere with the arteries. The only veins complicating interpretation of the images were those veins to the circle of Willis, such as the inferior sagittal sinus or the vein of Galen, when a lateral view was chosen for demonstration of the aneurysms. With all parameters correctly chosen (delay time, flow, window width 2700-3100HU, window centre 7501100 H U ) an almost selective good quality display of the brain stem arteries and the skull base could be achieved.
Aneurysms All nine aneurysms with a diameter between 5 and 28 mm could be identified correctly by two independent radiologists (N.H., K.N.); intra-arterial angiography, the reference study, was available in eight patients. Assessment of size, location, and the relation of the aneurysm to the parent vessel, with a broad base or narrow neck, was satisfactory. Figures 2-4 show representative examples. Figure 4 a demonstrates the effect of setting the window width and centre slightly too low: the image is too bright and the distinction between skull bone and vessels is impaired.
Arteriovenous malformations The diagnosis could be confirmed in agreement with intra-arterial study in all 13 patients with supratentorial AVMs.
With 3D reconstruction one gains useful information about the main feeding arteries, clearly identified in every patient. The nidus of the A V M could be seen in relation to the skull, providing a good orientation for the therapeutic radiologist because of the 3D impression. The site of dilated draining pial veins is also clearly shown. The radio-opaque embolisation material was well demonstrated in all six patients who underwent embolotherapy. Figures 5 and 6 show the representative images. After the postprocessing procedures, which usually take about 20 min, it is possible to rotate the A V M freely, to cut into the skull and the malformation from any desired direction and to use a simulated light source to produce the most helpful 3D projections (Fig. 6 c). With the stored SCT data file this procedure is repeatable as often as desired with different thresholds.
Discussion In our study of arterial SCT and 3D reconstruction we were able to confirm the diagnosis made on the intraarterial study in all patients except one who did not undergo DSA. A major part of the information necessary for adequate therapy can be gained with arterial SCT. Detailed demonstration of the vascular anatomy depends on the intravascular iodine concentration, the
Fig.5a-e A 44-year-old patient suspected of having an intracra- 9 nial haemorrhage, a There is fresh bleeding in the right hemisphere, b SSD image obtained after three consecutive SCT acquisitions and 3D reconstruction. The extent of the arteriovenous malformation (AVM), main feeding vessel (long arrow) and nidus (short arrow) are well shown. The blood does not mask the AVM as on MRA (ct. c). In this projection the draining veins cannot be seen. d SSD image without contrast medium after embolisation with histoacrylate. The radio-opaque embolisation material can be clearly identified (short arrow), e SSD image after embolisation, same projection: the decrease in size of the lesion is demonstrated
249
Fig.6a-c A 56-year-old patient with a left hemisphere AVM, investigated for severe headache, a The 3D projection "from below" shows dilated draining pial veins and running towards the superior sagittal sinus (short arrows). The main feeding vessel is a hypertrophied left M C A (long arrow). b Intra-arterial D S A for comparison, e Semifrontal projection: part of the frontal bone was "cut away", providing a better view of the dilated feeding artery (long arrow) and nidus (short arrow)
250
Fig.7 Traumatic caroticocavernous fistula. Coronal 3D cut through the anterior skull base/ facial bones. The importance of simultaneous display of the facial bones is evident: the arterialised superior ophthalmic vein lies directly below the orbital roof (arrow)
Fig.8a, b Colour-coded images of the aneurysms shown in Figs.2 and 4, created on a separate postprocessing computer. Better differentiation between bones and vessels is achieved
threshold selected and the voxel size. With a fixed threshold and voxel size in a 512 x 512 matrix a decreasing iodine concentration results in increasing partial volume effects, especially when the vascular luminal diameter becomes less than the section thickness [1]. With adequate intravascular iodine concentration, and delay and scanning times, the attenuation values of the contrast-enhanced vasculature are high enough to avoid the inclusion of masking parenchyma or a bleed. Over 95 % of intracranial aneurysms are within or close to the circle of Willis [2]. This region can be covered with a maximum of two SCT acquisitions with the maximum data provided by the minimum possible slice thickness and table speed. Although the spatial resolution of 0.5 mm (in-plane) is about one tenth of that afforded by intra-arterial DSA [3], it is high enough to depict arterial branches beyond the aneurysm (Fig. 4 a). The patient's head can easily be fixed and the examination is not dependent on any breath-holding efforts of patients who may be severely ill. The patients do not need to be moved once intracranial haemorrhage is shown by CT. SCT takes little additional time, the whole examination is faster than DSA and only minimally invasive. At the same time one gains information about the underlying cause of the bleed. Thus the arterial system of the skull base is highly amenable to examination by arterial SCT. With its postprocessing possibilities this new technique may afford better studies, especially of the base of the aneurysm, which may be broad or narrow. Colour-coded display of different ranges of attenuation values can give better distinction between contrast-enhanced vessels and bones. According to various studies, magnetic resonance angiography (MRA) reaches specificity and sensitivity of up to 95 % for intracranial aneurysms [4-8] even in patients with subarachnoid haemorrhage (SAH), as compared to intra-arterial DSA. The smallest aneurysm
251
detected by Gasparotti et al. [5] was 2.5 mm, probably below the critical size for detection by SCT. The m e t h o d of choice for the demonstration of intracranial h a e m o r r h a g e is CT. The more sophisticated the examination technique, the longer it takes. The high resolution M R A used by Gasparotti et al. [5] took nearly 20 rain and involved transport to the M R I unit. From our experience it is questionable whether an intubated, comatose patient tolerates this because managing such a patient in an M R I unit is difficult and time-consuming. The well-known limitations of arterial M R A include difficulties in correctly demonstrating vessels with flow turbulence (especially large vessels like the carotid arteries) or low-flow conditions [9]. With SCT it is not possible to achieve completely selective demonstration of the arteries as with intra-arterial studies but because the arteries of the circle of Willis are not accompanied by veins this disadvantage is of no great importance. With regard to the angioarchitecture of AVMs it is desirable to include the veins in the 3D image. The caudocranial projections permit reliable delineation of at least the main dilated draining pial veins and simultaneously of their relationship to the skull, which may be useful preoperative information (Fig. 6 a). It is therefore important to avoid a mistake we made at first: we did not include the whole skull in our measurements, thereby failing to show dilated draining pial veins at the vertex. Figure 7 underlines the possible usefulness of simultaneous display of the skull and facial bones in a patient with a traumatic arteriovenous fistula.
SCT fails to give information about the blood flow characteristics within the complex angioarchitecture of AVMs, as can be partly achieved by M R A . With M R A it is also possible to display the direction of blood flow and to estimate its [9, 10], e.g., high-flow venous drainage, which is especially important if embolotherapy is planned. SCT seems to have the same potential as M R A for showing the main feeding vessels, the angiomatous nidus and large draining veins. Some of the patients presented here also underwent MRA. Our personal impression in patients with a h a e m a t o m a due to an A V M was that, unlike SCT, M R A display of the A V M is compromised by the overlying signal of the blood. Once the therapeutic strategy of surgery, radiation or embolisation is established by intra-arterial DSA, arterial SCT with 3D reconstruction may serve as a minimally invasive, faster and less expensive [3] m e t h o d for follow-up, which may reduce the n u m b e r of invasive procedures. Patients at risk with arterial hypertension or clotting disorders especially may benefit. SCT is also an excellent m e t h o d for direct 3D demonstration of the radio-opaque embolisation material. The current limitations of arterial SCTwith 3D reconstruction in AVMs are its spatial resolution, which is not high enough to exclude the existence of additional small feeding vessels and the lack of information about the velocity of blood flow within the different parts of the AVM. Solutions, such as the attempt to create subtraction angiographic images are in progress, using advanced postprocessing possibilities.
References 1. Kalender WA, Vock P, Polacin A, Soucek M (1990) Spiral-CT: eine neue Technik for Volumenaufnahmen. R6ntgenpraxis 43:323-330 2. Jellinger K (1979) Pathology and aetiology of intracranial aneurysms. In: Pia HW, Langmaid C, Zierski J (eds) Cerebral aneurysms. Advances in diagnosis and therapy. Springer, New York Berlin Heidelberg, pp 5-19 3. Rubin GD, Napel S, Dake MD, Walker PJ, McDonnell CH, Marks MR Jeffrey RB (1992) Spiral-CT creates 3D neuro, body angiograms. Diagn Imaging 8: 6674 4. Futatsuya R, Seto H, Kamei T, Nakashima A, Kakishita M, Kurimoto M, Endoh S (1994) Clinical utility of threedimensional time-of-flight magnetic resonance angiography for the evaluation of intracranial aneurysms. Clin Imaging 18:101-106
5. Gasparotti R, Bonetti M, Crispino M, Pavia M, Chiesa A, Galli G (1994) Subarachnoid hemorrhage assessment in the acute phase with angiography, with high resolution magnetic resonance (angio-MR). Radiol Med 87: 219-228 6. Gouliamos A, Gotsis E, Vlahos L, Samara C, Kapsalaki E, Rologis D, Kapsalakis Z, Papavasiliou C (1992) Magnetic resonance angiography compared to digital subtraction angiography in patients with acute subarachnoid hemorrhage. Neuroradiology 35: 46-49 7. Horikoshi T, Fukamachi A, Fukasawa I (1994) Detection of intracranial aneurysms by three-dimensional time-offlight magnetic resonance angiography. Neuroradiology 36:203-207
8. Schuierer G, Huk WJ, Laub G (1992) Magnetic resonance angiography of intracranial aneurysms: comparison with intraarterial digital subtraction angiography. Neuroradiology 35:50-54 9. Ruggieri PM, Masaryk TJ, Ross JS, Modic MT (1992) Intracranial magnetic resonance imaging. Invest Radiol 27 [Suppl 2]: 33-39 10. FiJrst G, Bettag M, Fischer H, Skutta B, Hofer M, Steinmetz H, Kahn T (1993) The selective arterial and venous MR angiography of intracranial arteriovenous malformations. Fortschr R6ntgenstr 159:71-77