Normal Venous Anatomy of the Brain: Demonstration with Gadopentetate

Transcription

1 1107 Normal Venous Anatomy of the Brain: emonstration with Gadopentetate imeglumine in Enhaned 3- MR Angiography onald W. Chakeres 1 Petra Shmalbrok 2 Martha Brogan 1 Chun Yuan 2 Lane Cohen 1 This investigation evaluates whether gadopentetate dimeglumine enhanement of three-dimensional (3-) aquisition MR angiography an generate linially useful images of the normal venous anatomy of the brain. 3- MR angiography of normal erebral arterial anatomy has made rapid progress, although demonstration of detailed venous anatomy with similar tehniques has been muh less revealing. To overome the limitation of slow venous flow, IV gadopentetate dimeglumine ontrast enhanement was used to alter the relaxation times of blood, thus augmenting the venous signal. Several groups of patients were evaluated: we studied eight patients both with and without ontrast enhanement, 20 patients and volunteers with multiple tehniques to determine optimal tehnial parameters, and seven patients in whom enhaned MR studies were ompared with standard seletive biplane ut-film arterial angiograms. Only the large dural sinuses (suh as the transverse sinus) ould be seen on unenhaned studies owing to the saturation of slowly flowing venous spins. With ontrast enhanement, many of the important small and large erebral venous strutures were routinely seen with reasonable sanning times (7 min). The venous anatomy was well seen for approximately one-half hour after injetion and orrelated well with angiograms. There are several important limitations to this tehnique, inluding a limited field of view, variable visibility of speifi veins owing to tehnial and physiologi fators, onfusion of enhaning non-flow-related strutures, and lak of detailed physiologi information. Single exitation 3- MR angiograms are insensitive in the evaluation of erebral venous strutures. Enhanement with gadopentetate dimeglumine affords rapid sanning and exellent visualization of the pertinent venous anatomy. The best image quality was obtained with a sequene of 50/7/30 (TR/TE/flip angle). AJNR 11: , November/eember 1990; AJR 156: January 1991 Reeived August 29, 1989; revision requested November 17, 1989; final revision reeived June 1, 1990; aepted June 4, ' epartment of Radiology, ivision of Neuroradiology, Ohio State University Hospital, Ohio State University College of Mediine, 410 W. 10th Ave., Columbus, OH Address reprint requests to. W. Chakeres. 2 General Eletri Medial Systems, Milwaukee, Wl /90/ Amerian Soiety of Neuroradiology A number of MR angiographi tehniques have been proposed that use the advantages of flowing blood to differentiate moving from stationary spins. The two most ommon tehniques are based on time-of-flight effets of moving spins or on motion-indued phase shifts [1-7]. The most time-effiient methods are those that use three-dimensional (3-) volume and multiple thin two-dimensional (2-) [8, 9] time-of-flight tehniques, sine they use only single exitation data and there is no need to subtrat different data sets. Subtration of 3- MR angiograms with and without gadopentetate dimeglumine has been reported to improve visualization of the venous strutures (Seider M, unpublished data). Phase-ontrast MR angiegraphi tehniques evaluate differenes in phase indued by motion in the presene of magneti field gradients [5]. 3- volume phase-ontrast tehniques generally have long aquisition times, up to 40 min. The 3- time-of-flight volume MR angiographi tehnique has made rapid progress in the aquisition of images of normal erebral arterial anatomy, yielding good orrelation with standard angiography [1 0-13]; however, demonstration of detailed venous anatomy has been muh less revealing. Only large venous strutures suh as the dural sinuses are seen routinely, and even then, not to the best advantage.

2 1108 CHAKERES ET AL. AJNR :11, November/eember 1990 Noninvasive evaluation of the erebral venous strutures is desirable in order to ensure a omplete evaluation of the brain vasulature. Venous thrombosis, striture, ompression, displaement, and anomalies are all important angiographi findings that ordinarily are not visible on images designed to optimize single-exitation 3- volume MR angiographi arterial studies. The poor visualization of the venous anatomy with the 3- MR angiographi tehnique is partially a result of slower moving venous spins that are saturated within the imaging volume. To avoid this problem, we generated MR angiograms that were not based entirely on flow effets, but improved by relaxation time differenes between vasular strutures and stationary tissues. For this purpose, the ontrast between blood and brain tissue was enhaned by IV injetion of gadopentetate dimeglumine. Beause this investigation was primarily foused on development of the tehnique and so many different parameters were varied, this artile does not attempt to address the utility and auray of visualization of speifi veins and abnormalities. It is a study speifially to review the tehnial parameters that are optimal for gadopentetate-dimeglumine-enhaned venous MR angiography. Subjets and Methods Thirty-five patients and volunteers 4-75 years old (mean age, 45 years) were evaluated. Almost all of the patients were adults. Images were aquired on a 1.5-T General Eletri Signa system using a gradient-eho 3- volume imaging sequene, whih allows very short TEs (Shmalbrok P, unpublished data). The sequene uses firstorder flow-motion ompensation in both read- and slie-seletion diretions. The slie thikness an be varied from 0.7 to 5 mm, and the field of view from 16 to 48 em. Initially, the shortest possible TE (< 7 mse) required field-of-view restritions in the frequeny-enoded diretion. Later versions of the pulse sequene ode avoided this limitation. In most ases, mm-thik slies were aquired with a field of view of em and a 128 x 256 matrix resolution. Beause only a single aquisition is required, san times were typially under 7 min. Thirty slies ould be aquired in 3Y2 min. Angiographi displays were generated by using ray-traing methods on a stak of thin setions [5, 7, 14]. In the omputational algorithm, parallel rays through a hosen imaging volume seleted the points with maximum signal from the stak of individual images. In this fashion, a new projetion image was generated. By hanging the diretion of the parallel rays, projetion images of various angles ould be generated. A minimum of six different projetions were alulated for any hosen stak of slies. The projetions were reated to orrespond to standard anteroposterior and lateral angiegrams. We initially obtained unenhaned images to evaluate the visualization of venous strutures (Fig. 1). Then, we ompared eight MR angiographi studies with and without IV gadopentetate dimeglumine enhanement (1 0 ml, 4.7 g) in the same patient or volunteer to evaluate the visualization of erebral venous anatomy with enhanement (Fig. 2). Varying T1 ontrast weighting on unenhaned and ontrast-enhaned images was ahieved by manipulating the TR and flip angle. A 8 Fig. 1.-Effet of variable flip angle on unenhaned 3- MR angiography. A-, Lateral projetions from unenhaned axially aquired 3- volume images, 50/7 (TR/TE), with variable flip angles of 15 (A), 30 (8), 45 (C), and 60 (0). Pixel size was 1.5 x 1.8 x 0.9 mm. Flow ompensation was applied in both read- and slab-exitation diretions. On lower-flip-angle image, visualization of venous strutures, inluding transverse (solid blak arrows) and straight (open arrows) sinuses, is faint. With larger flip angle, straight and transverse sinuses are almost ompletely absent. Even arterial strutures (white arrows) are degraded by saturation effets.

3 AJNR:11, November/eember 1990 MR ANGIOGRAPHY OF CEREBRAL ANATOMY 1109 Fig. 2.-Multiple projetions with and without ontrast enhanement. 3- MR angiograms, 47!7 (TR/TE), were aquired sagittally and are displayed in multiple projetions (top left image is lateral; lower right is anteroposterior). Voxel size was 1.5 x 1.9 x 0.9 mm. A, Unenhaned study, with a 15 flip angle, poorly displays venous anatomy, although proximal arterial system is visible. B, Enhaned study, with a 30 flip angle for optimal results, shows a dramati improvement in visualization of many venous strutures. = arotid anal, sss = superior sagittal sinus, v = ortial vein, iv = internal erebral vein, iss = inferior sagittal sinus, a = arotid artery, ss = straight sinus. A B Spoiler gradients on all axes allowed the generation of images that were more T1 weighted. We varied the TR from 33 to 50 mse. Initially, unenhaned MR angiograms with flip angles of 15, 30, 45, and 60 were obtained in the same volunteer (Fig. 1 ). Then, similar series of images was obtained with gadopentetate dimeglumine in the same volunteer (Fig. 3). The TE was varied from 7 to 14 mse to evaluate its effet on visualization of the veins, as well as fat suppression due to hemial shift [15]. To evaluate washout of ontrast material over time, onseutive ontrast-enhaned images in the same volunteer were ompleted (Fig. 4). To evaluate the effet of setion plane and image volume on venous visualization, onseutive setion planes in the same volunteer were ompared after ontrast injetion (Fig. 5). A omparison was made of single- and multiple-image volume seletion for ray-traing projetions (Fig. 6) and the original images were reviewed for anatomi detail (Fig. 7). We diretly ompared standard seletive biplane ut-film erebral angiograms and MR angiograms from the same seven patients (Figs ). A general evaluation was made of the quality, extent of visualization of the venous system, artifats, and optimal imaging parameters with erebral venous MR angiography. Beause the examinations were done with many different tehniques, a detailed quantitative evaluation of speifi veins or abnormalities with varying tehniques was not possible. Results The unenhaned MR angiograms displayed the venous strutures poorly (Fig. 1 ). The images with higher flip angles further limited venous visualization. Comparison of enhaned and unenhaned sans of adult patients was done as soon as possible after injetion of gadopentetate dimeglumine. There was remarkable improvement in the visualization of the venous system with enhanement (Fig. 2). Most of the major venous strutures were well seen. Visualization of the arterial strutures was not signifiantly improved with ontrast material. Changes in T1 weighing did have a notieable impat on image quality in both enhaned and unenhaned examina-

4 1110 CHAKERES ET AL. AJNR:11, November/eember 1990 A B Fig. 3.-Variable-flip-angle ontrastenhaned MR angiograms. A-, Axially aquired images were obtained on two onseutive days to avoid hanges related to ontrast washout (A and 8, one day; C and, the next). The only aquisition parameter varied was the flip angle: 15 (A), 30 (8), 45 (C), and 60 (). Images are displayed as lateral angiograms. Image with 60 flip angle () shows extensive saturation of superior sagittal sinus and transverse sinuses in inferior portion of image (arrow). Other images are of omparable quality, but there are subtle differenes. Image with 15 flip angle (A) shows less detail of deep venous system (arrows). Original images showed Jess ontrast between gray and white matter strutures than did higher-flip-angle images. Choroid plexus is more prominent on image with 45 flip angle (C), leading to onfusion with flowing strutures. Enhanement of nasal muosa is visible on all images. A B Fig. 4.- Time-related ontrast washout. A-, Four onseutive sagittally aquired images, 50/11/30, at 4 (A), 17 (8), 31 (C), and 44 () min after injetion of 10 ml of gadopentetate dimeglumine. Voxel size was 1.5 x 1.6 x 0.8 mm. Only the entral 22 images of a total of 60 were postproessed for raytraing projetion images shown. Best venous image ontrast is seen on 4- min study (A). Midline venous strutures, inluding superior sagittal sinus (open blak arrows) and internal erebral veins (solid white arrows), are seen well (A). A number of important arterial strutures, inluding periallosal (open white arrow) and basilar (solid blak arrows) arteries, are seen on 4-min image also. With longer postinjetion delays, smaller vasular strutures are not seen as well (). Venous quality of 44-min image () is marginal. There is minimal hange in visualization of large arterial strutures (solid blak arrows) with time. Extensive enhanement of nasal muosa and wraparound artifat is noted (urved white arrows). tions. Varying the TR from 33 to 50 mse did not signifiantly alter the results exept for hanging the total examination time. Longer TR s were not onsidered beause of the prolonged examination time, and shorter TRs were not onsidered owing to tehnial limitations and expeted derease in signal to noise. On unenhaned studies with lower flip angles, only the large dural sinuses (suh as the superior sagittal and transverse sinuses) ould be seen (Fig. 1A). Owing to saturation of the slow-flowing venous spins, the larger the flip angle, the worse the venous visualization (Figs. 1 B-1 ). On

5 AJNR :11, November{eember 1990 MR ANGIOGRAPHY OF CEREBRAL ANATOMY 1111 Fig. 5.-Comparison of axial and sagittal aquisitions. A-, Contrast-enhaned images, 52/11{30, were obtained with both axially (A and 8) and sagittally (C and ) aquired tehniques, then postproessed with ray traing to generate similar anteroposterior (A and C) and lateral (8 and ) projetions. Eah aquisition has advantages for visualization of speifi vessels. eep venous system is better seen on axially aquired series (A and 8), inluding vein of Galen (urved white arrows), transverse sinus (straight white arrows), and internal erebral veins (open white arrow). On sagittally aquired images (C and ), superior sagittal sinus (open blak arrows), jugular fossa, and ortial veins (urved blak arrows) are seen best. A B J the ontrast-enhaned studies, large flip angles also led to more venous saturation (Fig. 3). Inrease in bakground signal from non-flow-related enhaning tissues (nasal muosa and brain) beame more pronouned with higher flip angles (Fig. 3). A flip angle of approximately 30 was optimal, though there was a wide range that was adequate. The lower-flipangle original images had less T1 weighting. The higher-flipangle original images were less attrative. The effet of the various TEs on image quality was important. No substantial signal loss for the venous strutures was observed when inreasing TEs from 7 to 15 mse. Modifiation of the TE between 7 and 11 mse, however, aused a variation in the bakground signal intensity of the fatty tissues (Fig. 7). Best results were obtained with ate of 7 or 11 mse, where the magnetization of fat and water aligns in an antiparallel fashion, while a TE of mse with parallel magnetization inreased the bakground signal [15]. This was important in the suppression of unwanted signal from the salp on the ray-traing projetion images. The improvement in venous visualization lasted approximately 30 min after injetion (Fig. 4). The sooner the examination followed the injetion, the better the result, but the ontrast,load was adequate for at least two aquisitions. Orientation of the aquisition setion plane, slie thikness, and total 1\lUmber of slies were important fators, sine they defined thf'l total imaging saturation volume. When the imaging volume ompletely enompassed the majority of a venous territory, the results were ompromised (Fig. 5). This indiates that the signal intensity is not entirely due to T1 ontrast, but is also dependent on inflow effets. For example, rostral axial setions of the vertex did not demonstrate the superior sag- ittal sinus as well as a diret sagittal series did (Fig. 5A). However, the deep venous system was usually seen better on axial images that did not inlude the vertex of the brain than on diret sagittal images (Fig. 58). Coronal aquisitions, though possible, were not frequently used sine they did not orrespond with the venous anatomy in terms of overing a wide area. The angiographi ray-traing projetion generated from smaller proessed volumes best demonstrated smaller strutures (Fig. 6) [14]. The larger strutures (for example, the superior sagittal sinus) were less sensitive to the size of the reonstruted volumes. The original slie images were of good quality, with T1- weighted ontrast similar to that of spin-eho images (Figs. 7 and 11 ). iagnosis of soft-tissue disease was possible (Figs. 7 and 11 ). The images were not learly better than standard spin-eho images; however, the thin slie thikness was advantageous in the visualization of small strutures. Comparison of the MR and standard angiograms onfirmed that, on the enhaned images, many of the important erebral venous strutures were seen routinely. In most instanes, MR angiography orrelated well with standard seletive erebral angiography (Figs ), although the results varied among patients. Routinely visualized strutures inluded superior and inferior sagittal sinuses (Fig. 2), transverse and sigmoid sinuses (Fig. 11 ), vein of Galen (Fig. 9), basal vein of Rosenthal (Fig. 9), internal erebral and thalamostriate veins (Fig. 8), lateral mesenephali and jugular veins (Fig. 5), and ortial veins (Figs. 5 and 6). Oasionally, the septal (Fig. 6), preentral erebellar, and other smaller veins were visible. Contrast enhanement helped define the extent of disease in

6 1112 CHAKERES ET AL. AJNR :11, November/eember 1990 A 8 Fig. 6.-Effets of ray traing postproessing. A-, Four images generated from one set of ontrast-enhaned sagittally aquired images, 52/11/30, show effets of different postproessing tehniques on visualization of various venous strutures. A is single midline setion from imaging blok. Visualization of even small veins, suh as internal erebral and septal veins (straight solid arrows), is superb. B was proessed from entral four images. C and were proessed from 22 and 60 images, respetively. Most smaller midline veins are not visible on thiker slabs (C and 0). On large imaging blok (0 ), ortial veins (urved blak arrows), omplete superior sagittal sinus (open arrows), and jugular fossa (urved white arrow) are better displayed than on narrow midline image. A 8 Fig. 7.-Comparison of spin-eho and 3- images. A and B, Gadopentetate-dimeglumine-enhaned axial images in patient with aousti neuroma. Spin-eho image, 600/30, with a 256 x 256 matrix and 5-mm-thik setion (A) and single 1.5-mm 256 x 128 setion from a SOslie 3- MR angiogram, 47/11/30 (8), were ompared. Both show enhaning intraanaliular mass on left (urved arrows). MR angiogram (B) shows high signal from arotid arteries (open arrows) and sigmoid sinuses (solid straight white arrows). Blak interfae of sub-temporal fossa fat and adjaent musles is seen seondary to hemial-shift effet (blak arrows).

7 AJNR :11, November/eember 1990 MR ANGIOGRAPHY OF CEREBRAL ANATOMY 1113 Fig. B.-iret omparison of anteroposterior standard and MR angiegrams. A-, Standard ut-film seletive bilateral anteroposterior ommon arotid angiograms (A and 8) and axially aquired MR angiograms, 40/7/30, raytraing projetions (C and ). Only entral volume of brain surrounding deep venous system was proessed on C to better display deep venous anatomy. Complete volume was proessed on. An enhaning suprasellar raniopharyngioma is seen (straight arrows, ). Mild subfaline herniation to left is seen by distortion of internal erebral and thalamostriate veins (solid arrows, A C). MR angiogram (C) aurately reflets loal distortion. Complete data reonstrution on shows superior sagittal sinus and middle erebral arteries (urved arrows) better. Choroid plexus of lateral temporal horns is seen also (open arrows, C). A B ases of an arteriovenous malformation (Fig. 1 0), aousti neuroma (Fig. 7), and metastati tumor obstruting the jugular vein (Fig. 11). isussion Most MR angiographi tehniques exploit one of two strategies to obtain sensitivity to flowing blood. The first strategy relies on hanges in the phase of the transverse magnetization, whih are introdued when blood moves in the presene of magneti field gradients. Those spins moving in the diretion of inreasing gradient strength will advane in phase, while those moving in the opposite diretion will fall behind the phase of stationary tissue. Phase-sensitive tehniques usually involve subtration of data obtained with different gradient shemes, suh as bipolar gradient pairs [4], with and without first-order motion ompensation [3], or use of different first-moment magneti field gradients (Pel NJ, unpublished data). Investigation of omplex blood flow requires aquisition of three separate data sets for the three gradient diretions, thus leading to long examination times (up to 40 min). Sine the hanges in the phases are diretly related to the veloity of flow, physiologi information an be obtained. Phase-sensitive tehniques are apable of deteting relatively slow flow over large fields of view, sine they are less sensitive to saturation. They are sensitive only to the omponent of flow oinident with the diretion of the applied magneti field gradient. For routine linial studies they are very slow, partially beause multiple aquisitions are needed. Beause of the multiple aquisitions, high-resolution studies take even longer. We initiated our study to find tehnial parameters that would eliminate as many of these limitations as possible.

8 1114 CHAKERES ET AL. AJNR:11, November/ eember 1990 A 8 Fig. 9.-iret omparison of a lateral standard and MR angiograms. A, Standard ut-film seletive arotid lateral angiogram shows muh of deep venous system. B, Corresponding axially aquired MR angiogram, postproessed lateral projetion. eep venous system is well seen on both studies. Carotid arteries are well seen. VA= venous angle, ICV =internal erebral vein, VG =vein of Galen, BVR = basal vein of Rosenthal, SSS = superior sagittal sinus, SS = straight sinus, T = torular Herophili. The seond MR angiographi strategy relies on time-offlight effets to transport flowing spins with a different RF exitation history than the stationary tissue into the imaging region. The different exitation history an onsist of seletive presaturation [16] or inversion pulses (Wright GA, unpublished data), whih are used to "tag" the signal of inflowing spins. Another time-of-flight tehnique takes advantage of the inflow of "fresh " unsaturated spins in a volume of saturated stationary spins, as in 2- and 3- time-of-flight angiography. Aquisition of a single 2- time-of-flight MR angiogram that highlights the signal from moving spins is possible in just a few seonds. The diretion of flow should be perpendiular to the setion plane for ideal results. Marked differenes in signal between moving spins and stationary tissue are ommon with this tehnique. Flow parallel to the setion plane an indue saturation. The original 2- MR angiograms are of poor quality with respet to anatomi detail, sine the signal of the stationary tissue has a poor signal-to-noise ratio. Also, the resolution (slie thikness) is limited by gradient strength. To reate a ray-traing projetion, multiple images are needed, inreasing the aquisition time. Motion between setions an degrade image quality [9], similar to CT postproessing artifats. The time-of-flight methods also are limited to the investigation of relatively fast flow (arteries or highveloity veins) andjor small imaging regions, sine the inflow of initially "fresh" spins is rapidly saturated as the imaging volume is traversed. For example, slow moving spins in the internal erebral veins would be more saturated with sagittal than with oronal images. These differenes an produe regions of low signal that are not related to a true stenosis. Beause of the limitations of 2- MR angiographi tehniques, we deided to investigate the visualization of the venous anatomy with our tehniques. Other investigators have reported early results on the use of gadopentetate dimeglumine with 3- MR angiographi time-of-flight imaging tehniques (Seiderer M, unpublished data). They subtrated two separate series of 3- MR angiegrams before and after ontrast enhanement. It was found that visualization of the venous system was partiularly improved. It was also reported that the anatomi information obtained with this tehnique was omparable to that obtained with digital subtration angiography. The optimal tehnial parameters were similar to those we onluded to be optimal, 33/11/40 (TR/TE/flip angle); however, their tehnique has several major drawbaks: patient motion during or between the two series seriously degraded image quality, the aquisition time was at least twie that of a single aquisition, and eah aquisition took 22 min. We wished to develop a linially useful single fast-aquisition MR angiographi tehnique that would provide highresolution quality on the original and ray-traing images. It was lear from prior experiene with T1-weighted spin-eho imaging that the veins enhaned intensely following injetion of ontrast material. We attempted to take advantage of this phenomenon with a 3- MR angiographi tehnique. Our approah an be lassified as a hybrid time-of-flight T1- weighted method. We use a 3- pulse sequene variation that ould be used for arterial studies, but it is more T1 weighted. The 3- tehniques inlude high resolution with slie thikness potentially thinner than with 2- MR angiography, short aquisition times (30 setions in 3112 min), and exellent original T1-weighted images owing to the improved signal-to-noise ratio of the 3- MR angiographi tehnique. This study demonstrates that unenhaned single-exitation 3- MR angiography is an insensitive examination for demonstrating venous anatomy (Fig. 1 ). The venous flow rates in even the large dural sinuses are muh slower than the arterial rates; thus, the spins reah saturation before they exit the

9 AJNR :11, November/eember 1990 MR ANGIOGRAPHY OF CEREBRAL ANATOMY 1115 Fig. 10.-Comparison of standard and MR angiograms in patient with frontoparietal arteriovenous malformation (AVM) demonstrates importane of ontrast enhanement. A and 8, Lateral (A) and anteroposterior (8) late arterial right arotid artery angiograms show large AVM (open arrows). Large arterial feeders (long thin arrows). Huge dilated draining veins (shorter straight arrows) are seen oursing superiorly to empty into superior sagittal sinus (urved arrows). C and, Unenhaned 3- MR angiograms, 47/11/30, lateral (C) and anteroposterior () ray-traing images, display enlarged arterial feeders from anterior and middle erebral arteries (arrows) toward AVM. AVM and its large draining veins, however, are not visible. E and F, Contrast-enhaned 3- MR angiograms, 47/11/30, lateral (E) and anteroposterior (F) ray-traing images, better display large luster of draining veins (solid white arrows). Without ontrast enhanement, omplete extent of abnormality would be underestimated. A B E F imaging slab. This is partiularly the ase for thik-volume slabs, short TRs, and higher flip angles (Fig. 1 ). Very little signifiant linial information relating to the venous system was generated from 3- MR angiographi tehniques used for optimal arterial visualization. The limited visualization of the veins is not a disadvantage if the goal is to display the arteries only. To overome the limitations of unenhaned single-exitation 3- MR angiography, we demonstrated that ontrast enhanement markedly improves venous visualization by altering the relaxation times of venous blood rather than relying solely on rapid flow veloities (Figs ). Beause an intat blood-brain barrier exludes ontrast material from the brain, there is limited enhanement of the brain itself (Figs. 2 and 3). The ontrast agent was onentrated in the blood pool, ausing large hanges in the relaxation times of the venous blood in omparison with the surrounding brain. This tehnique may not be useful outside of the brain. The image ontrast for arterial strutures was not signifiantly altered with gadopentetate dimeglumine, sine the signal generated by rapid flow-related enhanement alone produes nearly maximum ontrast (Fig. 2). The degree of T1 weighting was an important fator affeting venous visualization. The short T1 relaxation time of blood following enhanement allows for more rapid reovery of longitudinal magnetization and, therefore, inreased signal intensity on T1-weighted images. Consequently, better differentiation of brain tissue from veins is expeted when the flip angle is inreased, up to a point. This was observed in the images with flip angles (Figs. 1 and 3). Inreasing the flip angle to 60 dereased venous image quality, primarily owing to large saturation losses of the flowing spins. There were also larger signal ontributions from other tissues, suh as fat and ontrast-enhaning strutures (nasal muous), obsuring the veins. A flip angle around 30 was found to be optimal for both the original and projetion images. The hoie of TE was important sine it indiretly affeted the overall quality of the images. Changes in the TE did not

10 1116 CHAKERES ET AL. AJNR:11, November/eember 1990 A 8 Fig. 11.-Comparison of standard and MR angiography of jugular olusion in elderly man with a history of multiple left lower ranial nerve palsies, suggesting a jugular fossa lesion. Biopsy showed metastati small ell arinoma. A, Anteroposterior digital subtration angiogram from late venous phase of aorti arh injetion shows omplete olusion of left jugular fossa and vein (short white arrows). Open arrow= superior sagittal sinus. B, Anteroposterior ray-traing image from 3-0 MR angiogram, ontrast-enhaned axially aquired series, displays patent right jugular bulb and vein (long white arrows) and normal arotid arteries (urved arrows) with omplete olusion of left jugular vein (shorter white arrows). Open arrow = superior sagittal sinus. C and, Axial slies from same series show tumor invading left jugular fossa (arrowheads) without involvement of adjaent brain or arotid arteries (urved arrows). Straight arrows= jugular veins. diretly hange the visualization of the venous strutures, but rather they affeted the regions ontaining fat and water (Fig. 7) [17]. TEs of 7 or 11 mse helped to suppress the fat signal, whih an be a serious problem, partiularly on ray-traing images of the vertex of the skull. A TE of 7 or 11 mse orresponds with the antiparallel alignment of the fat and water magneti vetors. For further image improvement it might be neessary to implement fat suppression in the image sequene. The TE also may affet image ontrast by dephasing effets, with potential loss of signal with longer TEs. Sine image ontrast is dominated in part by T1 effets, and sine the venous flow is slow, signal loss due to dephasing for larger voxel sizes or longer TEs did not pose a signifiant problem, in ontrast to arterial blood flow. For fast flow, larger voxel sizes yield lower signal, owing to the fat that the veloity distribution within the voxel auses spin dephasing. For more T1-based images and slow flow, however, larger voxels may yield improved signal to noise, and thus ontrast to noise. Beause the veloities of venous flow are slower than those of arterial flow, less higher-order motion is expeted. Beause of this, very short TEs are not as ritial as they are with fast arterial flow to prevent spin dephasing [1]. On the other hand, the use of gadopentetate dimeglumine dereases not only T1 but also T2 relaxation, neessitating an appropriately short TE to prevent potential signal loss from the vasular strutures. The physiologi exretion and redistribution of the gadopentetate dimeglumine ontrast material did have a profound effet on image quality. The 3-0 MR angiographi ontrastenhaned study must be ompleted rapidly after injetion, sine a high onentration in the blood pool is ruial to ensure a good study (Fig. 4). The quality of the venous anatomy is maintained for a relatively long interval after injetion (approximately 30 min). Repeat injetions of ontrast material may also be used to ensure adequate ontrast enhanement, as long as the total dose does not exeed reommendations. Physiologi time-of-flight effets of both the venous and arterial vessels are very important. Even with ontrast enhanement, influx of unsaturated spins is a omponent of the signal harateristi. Choie of the overall imaging volume in relation to normal venous flow is ruial beause of saturation effets. There was gradual loss of signal from the sagittal, transverse, and sigmoid sinuses and jugular bulbs on the more inferior setions of axially aquired images (Figs. 3, 9, and 11 ), seondary to progressive saturation of venous spins flowing into the inferior parts of the image volume, sine the spins remain in the exitation volume for an extended time. Another example of this saturation phenomenon is the fat

11 AJNR :11, November/eember 1990 MR ANGIOGRAPHY OF CEREBRAL ANATOMY 1117 that the jugular fossae were better seen on sagittally than on axially aquired images beause of inflow of spins from the transverse sinuses (whih are outside the sagittal exitation region) (Fig. 5). The same phenomenon may aount for better visualization of the deep veins on axially aquired rather than sagittally aquired images, in whih ase the superior portion of the brain is not inluded in the imaging volume, and thus allows for inflow of unsaturated spins (Figs. 5, 8, and 9). We found that a total ombination of the imaging volume, diretion, and veloity of venous flow has signifiant effets on visualization of the small and large veins. Review of the originally aquired and postproessed images is also essential, sine nonvisualization of veins on the projetion images may be artifatual and tehnique dependent rather than a true finding of olusion or other disease (Fig. 6) [14]. Beause the stationary bakground tissue has relatively high signal, a vein may well be visible on the original image but not evident on the ray-traing projetion (Fig. 6). The orrelation between the ut-film angiograms and the MR angiograms is quite ompelling. For example, the major dural sinuses are well seen on MR angiography without the problem of partial venous filling seen on the seletive angiegrams (Figs ). An aorti arh injetion or a diret venous atheterization is frequently needed for more omplete venous angiographi filling (Fig. 11). The deep venous system is also well seen on MR angiography, whih has lower resolution than standard angiography, but the 3- display and the ability to selet the proessed volume proved to be a major advantage. For example, if the dural sinuses obsure a small vessel on the standard angiogram, it may not be resolved (Fig. 8). On MR angiography, a rotated or regional volume view may aurately display the small vessel. Limitations inherent to the maximum-intensity ray-traing tehnique aounted for nonvisualization of small vessels (suh as the septal veins), whih were seen on the original images but not on the angiographi projetions of large volumes (Fig. 6). The smaller the ray-traing volume, the more detailed the depition of the enhaning strutures. Trimming the volume down ould exlude regions that were not of interest or that were troublesome, suh as the enhaning muosa of the sinuses. The quality of the original brain images alone is very good, so this MR angiographi tehnique ould substitute for other T1-weighted spin-eho images, saving time (Fig. 8) [17]. The ontrast is similar to that of T1-weighted spin-eho images. There is some loss of signal from magneti suseptibility differenes seen at bone, air, and brain interfaes, but the hanges were not severe sine the TEs were relatively short. Clear definition of ontrast-enhaning disease was possible (Figs. 7, 10, and 11). This MR angiographi tehnique has several major limitations. The 3- tehnique is not ideal for visualization of the omplete erebral venous system. The present imaging volume is inadequate to inlude the omplete head, and saturation is a major problem (Figs. 2 and 5). With this tehnique it would be neessary to aquire at least two different imaging volumes in whih the venous territory is not ompletely enompassed to allow inflow of unsaturated spins. This is not as serious a limitation with 2- time-of-flight tehniques, sine the imaging slie is small and there is little saturation effet. Also, saturation is not as serious a problem with phaseontrast imaging, where the omplete head an be studied in one series. Confusion of high-signal regions due to enhanement, other high-signal strutures, and blood flow is a potentially serious problem (Figs. 3, 7, and 8) [18]. For example, an enhaning tumor may obsure a vasular stenosis, sine the lesion may envelop the narrowed vessel, as when a superior sagittal sinus thrombosis is obsured by an enhaning fal x meningioma. Confusion an also be aused by enhanement of normal tissues, inluding the dural strutures, horoid plexus, nasal muosa (Fig. 4), and other vasular strutures without blood barriers (suh as the pituitary stalk). Other high-signal tissues unaffeted by ontrast enhanement (suh as salp fat or subaute hemorrhage) also may be a problem [18]. An enhaning tumor in the loation of normal venous hannels ould produe potentially onfusing findings. Review of spineho or other images usually resolves the potential misinterpretation. Charaterization of flow properties suh as diretion, quantitative veloity, and higher-order motion ontributions is not obtainable with this tehnique. This is a disadvantage if physiologi information is desired. Possible indiations for venous MR angiography inlude dural sinus thrombosis, venous angioma, varix, unexplained erebral hemorrhage, arteriovenous malformation, venous anomaly, jugular paraganglioma, giant jugular bulb, ontraindiation to routine angiography, and demonstration of the normal venous anatomy for surgial planning. Complete assessment of both the venous and the arterial anatomy is important. If MR angiography is to be used for more than just a rough sreening examination, extensive and onsistent visualization of the entire erebral vasular system is neessary. In summary, unenhaned single-exitation 3- MR angiography is quite insensitive in the evaluation of erebral venous strutures. Contrast enhanement allows for rapid sanning times and exellent visualization of most of the important venous anatomy. We doumented good orrelation of utfilm angiograms and MR angiograms. This tehnique ould be used as a venous sreening examination and ould also substitute for a T1-weighted enhaned study so that the total examination time would not be seriously lengthened. We found the best image quality was obtained with a 50/7/30 sequene. REFERENCES 1. Nishimura G, Maovski A, Pauly JM. Considerations of MRA by seletive inversion reovery. Magn Reson Med 1988;7: umoulin CL, Cline HE, Souza SP, Wagle WA, Walker MF. 3- time of flight MRA using spin saturation. Magn Reson Med 1989;11 : Laub GA, Kaiser WA. MR angiography with gradient motion refousing. J Comput Assist Tomogr 1988;122: umoulin CL, Hart HR. Magneti resonane angiography. Radiology 1986;161 : umoulin CL, Souza SP, Walker MF, Wagle W. Three-dimensional phase ontrast angiography. Magn Reson Med 1989;9: Ruggieri PM, Laub GA, Masaryk TJ, Modi MT. Intraranial irulation: pulse sequene onsiderations in three-dimensional (volume) MR angiography. Radiology 1989;171: Masaryk TJ, Modi MT, Ross JS, et al. Intraranial irulation: preliminary

12 1118 CHAKERES ET AL. AJNR :1 1, November/eember 1990 linial results with 3- (volume) MRA. Radiology 1989;171 : Gullberg GT, Wehrli FW, Shimakawa A, Simons MA. MR vasular imaging with a fast gradient refousing pulse sequene and reformatted images from transaxial setions. Radiology 1987;165: Keller PJ, rayer BP, Fram EK, Williams K, umoulin L, Souza SP. MR angiography via 2- aquisition and three-dimensional display. Work in progress. Radiology 1989;173: Hale J, Valk PE, Watts JC, et al. MR imaging of blood vessels using three-dimensional reonstrution: methodology. Radiology 1985;157: Wedeen VJ. Mueli RA, Edelman RR, Frank LR, Brady T J, Rosen BR. Projetive imaging of pulsatile flow with magneti resonane. Siene 1985;230: Masaryk T J, Ross JS, Modi MT, Lenz GW, Haake EM. Carotid bifuration: MR imaging. Radiology 1988;166: Tsuruda J, Halbah VV, Higashida RT, Mark AS, Hieshima GB, Norman. MR evaluation of large intraranial aneurysms using ine low flip angle gradient-refoused imaging. AJNR 1988;9: Anderson CM, Saloner, Tsuruda JS, Shapeero LG, Lee RE. Artifats in maximum-intensity-projetion display of MR angi0grams. AJR 1990; 154: Wehrli FA, Perkins G, Shimakawa A, et al. Chemial shift indued amplitude modulations in images obtained with gradient refousing. Magn Reson Imaging 1987;5: Edelman RR, Mattie HP, Kleefield J, et al. Quantifiation of blood flow with dynami MR imaging and presaturation bolus traking. Radiology 1989;171 : Steinberg PM, Ross JS, Modi MT, Tkah J, Masaryk TJ, Haake EM. The value of fast gradient-eho MR sequenes in the evaluation of brain disease. AJNR 1990;11 : Yousem OM, Balakrishnan J, ebrun GM, Bryan RN. Hyperintense thrombus on GRASS MR images: potential pitfall in flow evaluation. AJNR 1990;1 1 :51-58