MAGMA Magnetic Resonance Materials in Physics, Biology and Medicine

ELSEVIER

Magnetic Resonance Materials in Biology, Physics and Medicine 6 (1998) 62-69

High-resolution venography of the brain using magnetic resonance imaging Jiirgen R. Reichenbach a,., Marco Essig b E. Mark Haacke c, Benjamin C. Lee d, Christian Przetak a, Werner A. Kaiser ", Lothar R. Schad b a Institute of Diagnostic and Interventional Radiology, Friedrieh-Schiller University, Baehstrasse 18, 07740 Jena, Germany b Department of Radiological Diagnostics and Therapy, German Cancer Research Center, Heidelberg, Germany c Mallinckrodt Institute of Radiology, Washington University, St. Louis', MO, USA d Department o f Pediatrics, St. Louis Children's Hospital, Washington University, St. Louis, MO, USA

Received 1 December 1997; accepted 23 February 1998

Abstract

The purpose of this study was to evaluate a non-flow related magnetic resonance imaging method to visualize small veins independent of arteries in the human brain. A long TE, high-resolution 3D gradient echo MR acquisition was used to highlight venous information. The method is based on the paramagnetic property of deoxyhemoglobin and the resulting phase difference between veins and brain parenchyma at long echo times. The MR magnitude images were masked with a phase mask filter to enhance s/-nall structure visibility.. Venous information down to sub-pixel vessel diameters of several hundred microns is visible. Venous data are displayed in an angiographic manner using a minimum intensity projection algorithm. Both superficial veins and deep white matter veins are visible. The method has been successfully applied in volunteers. Preliminary results in patients with cerebral arteriovenous malformations indicate its potential in clinical applications. The proposed method is easy to implement and does not require administration of a contrast agent or application of specially designed rf pulses to highlight the veins. Rather it exploits the intrinsic magnetic properties (BOLD-effect) and the prolonged T* of venous blood. The method may be of diagnostic potential in the assessment of arteriovenous malformations or other vascular venous lesions. 9 1998 Elsevier Science B.V. All rights reserved. Keywords: MR angiography; MR venography; Deoxyhemoglobin; Brain

1. Introduction

In the past decade there has been a tremendous progress in the development of magnetic resonance angiography (MRA) for depicting vessels and blood flow as well as for evaluating vascular diseases. N o w a days, it is possible to cover extensive regions of vascular anatomy within a single breathhold [1]. However, to date, most of the advances have been made for the development of arterial magnetic resonance angiography. On the other hand, M R imaging of the venous * Corresponding author: Tel.: + 49 3641 935357; fax: + 49 3641 936767; e-mail: [email protected]@na.de

system of the brain is also important and m a y offer new and helpful insights into venous lesions and diseases [2-41. There exist several M R methods to evaluate the venous vasculature in the brain, such as 2D or 3D- time of flight (TOF) methods, 2D- or 3D-phase contrast M R A with low velocity encoded gradients (VENC), contrast enhanced M R A , or M P - R A G E subtraction venography [5-7]. Other methods utilize the dependence of blood's T2 on the time between 180 ~ re-focusing pulses in a spin-echo sequence to image blood while suppressing signal from all other tissues using a flowcompensated, water selective, short TI inversion recovery sequence with a long echo time [8], or applying

1352-8661/98/$ - see front matter @ 1998 Elsevier Science B.V. All rights reserved. PII S1352-8661(98)00011-8

J.R. Reiehenbaeh et al. / Magnetic Resonance Materials. in Biology, Physics and Medicine 6 (1998) 62-69

63

Fig. 1. Influence of the number of phase mask filter multiplications on the final projected venograms. Each minimum intensity projection (mIP) corresponds to the same targeted volume of 16 mm. Sequence parameters are: TR = 80, TE = 40 ms, c~= 25~ TH = 2 mm, 16 partitions, FOV = 180 • 240 mm2, matrix size 384 x 512.

specially tailored rf pulses to visualize primarily venous structures [9]. Recently, we have proposed a new, three-dimensional, long TE, flow-independent gradient-echo approach based on the B O L D (blood oxygenation level dependent) effect [10] to visualize and highlight venous structures in the sub-millimeter regime without application of an exogeneous contrast agent [11]. The purpose of this study was to further investigate the usefulness of the method, to determine different M R sequence parameters for optimisation of the venograms and t o apply the method in clinical cases of arteriovenous malformations to assess its clinical potential.

2. Materials and methods All data were acquired on a 1.5 T Vision s y s t e m (Siemens, Germany) using a conventional CP head coil. Written consent was obtained from the volunteers and patients participating in this study.

As has been described in detail in Reichenbach et al. [11], images were acquired with a first-order flow compensated, strongly T*-weighted, low bandwidth (78 Hz/ pixel), 3 D - F L A S H sequence. Gradient m o m e n t nulling was incorporated in all three image directions to reduce signal loss and to avoid displacement artifacts due to oblique in-plane flow. Sequence parameters were varied in a systematic way with echo times TE ranging from 30 to 60 ms, tip angles a from 15 to 35 ~ and matrix sizes from 256 to 1024. The F O V was held constant resulting in a minimum in-plane pixel size of 0.23 x 0.23 ram. Using a sagittal localizer image, typically 16-24 transverse M R partitions with 2 m m thickness were acquired. Magnitude and phase images were reconstructed from the complex raw data sets. Phase images are displayed with values between - 4 0 9 6 and + 4095 corresponding to - 1 8 0 a n d + 180 ~ respectively. The phase-contrast of venous vessels appears dark (with negative values) on these images due to the fact that paramagnetic structures have a higher resonance frequency, i.e. a stronger local magnetic field is

64

J.R. Reichenbach et al. / Magnetic Resonance Materials in Biology, Physics and Medicine 6 (1998) 62-69

present at these proton sites. On the other hand, white matter and cerebrospinal fluid have small positive values. To create the high-resolution venograms, the grayscale of the magnitude images was inverted to highlight the small dark venous structures while reversing all bright signal structures in the images. In a second step a maximum intensity projection (MIP) was then performed across these images. Conceptually, the whole procedure represents a minimum intensity projection (mIP). Targeted volumes of typically 5-10 slices, i.e. a subset of the total number of acquired partitions, were chosen for this mIP operation. An increase of the visibility of the venous vessels was achieved by constructing a phase mask filter image from the reconstructed phase images for each partition and multiplying these phase mask filters with the corresponding magnitude images of the 3D slab prior to the projection. The phase mask filter images are implemented by setting all positive phase values (between 0 and 180~ in each individual phase image to 1, whereas all negative phase values, ranging from - 180 to 0% are linearly scaled between 0 and 1. Consequently, the mask filter image contains only values ranging from 0 to 1, depending on the original phase value. This phase mask filter image can be multiplied several times with the magnitude image to increase contrast between veins and brain parenchyma. Grayscale values of pixels in the magnitude images containing venous structures will be selectively diminished because they are multiplied with a numerical constant smaller than 1, whereas the pixel values of other structures in the magnitude image having positive phase values, such as white matter or CSF, will not be changed. Depending on the number of performed multiplications the degree of diminution can be controlled which, in turn, will influence the created contrast. Intensity projection over targeted volumes as described above will further enhance the venous structures. Further details of the method can be found in Reichenbach et al. [11]. In an ongoing study, nine patients with angiographically proven cerebral arteriovenous malformations were enrolled in this study. Five patients were examined prior to single high dose radiation therapy and four were examined between 6 weeks and 3 months after the radiation therapy. In additon to the conventional clinical protocol, the MR venography sequence was run before application of contrast agent. The clinical protocol comprised pre- and post-contrast Tl-weighted images, Tz-weighted images and fluid-attenuated IR (FLAIR) images. A 3D, velocity-compensated FISP (fast imaging with steady-state precession) MRA sequence (TR/TE = 37/6.5 ms) with variable flip angle (TONE) and magnetization transfer contrast (MTC) was also used.

3. Results

Fig. 1 demonstrates the influence of successive multiplications of the magnitude images with the phase mask filter images prior to the intensity projection on the appearance of the final mIPs in a volunteer. The contrast between the venous vessel system and the brain parenchyma increases with increasing number of multiplications. The same holds true for the apparent connectivity of the venous vessel net, especially for the small deep white matter venous vessels around the ventricles. Varying the number of phase mask filter multiplications makes it possible to tailor the final mIP in a very fine and subtle way. Empirically, we have obtained the best results with 3-5 multiplications, depending on the echo time TE, number of partitions projected and the location of the 3D slab in the brain. Figs. 2 and 3 illustrate the influence of the echo time TE and the tip angle ~ on the final projected venograms, respectively. With increasing echo time, the contrast between brain and veins becomes sharper and the vascular system is delineated much clearer. This is due to the fact, that at longer echo times the signal cancellation between brain parenchyma and venous blood in a single voxel is more effective, leading to an overall smaller resulting voxel signal. However, at the same time the undesirable signal loss due to other static magnetic field inhomogeneities becomes more and more severe. This is clearly seen in Fig. 2 in the projected venograms, especially at TE = 60 ms, where considerable signal cancellation is observed in the anterior part of the brain due to the presence of the air/tissue interface between parenchyma and frontal sinuses. Most of the data were thus acquired with echo times of 40 or 50 ms. As expected, the T1 contrast between the venous vessels and different brain structures depends on the value of the tip angle. A tip angle of about 25 ~ represents a good compromise for depicting vessels and brain structure simultaneously with good contrast while keeping all other parameters constant. Fig. 4 demonstrates the importance of high spatial resolution for the success of the venography method. Shown in the figure are sections of the same targeted volume acquired with different matrix sizes of 256, 512 and 1024. The nominal in-plane resolution corresponds to 0.94 x 0.94, 0.47 x 0.47, and 0.23 x 0.23 mm 2, respectively. Although there exists some variability in small vein visibility between the images, it is evident that, in general, the small vessel details are better and more continuously delineated with the highest spatial resolution. There is evidence that the smallest vessels visible in the image have a diameter of about 100-200 /tin. The portrayed anatomical information agrees well with the known venous structures [12]. Clearly, a low resolution scan with a matrix size of only 256 is totally inadequate to pick up, for instance, the fine details of

J.R. Reichenbach et a l . / Magnetic Resonance Materials in Biology, Physics and Medicine 6 (1998) 62 69

65

Fig. 2. Influenceof echo time TE on the projected venograms. Each minimum intensity projection corresponds to the same targeted volume of 16 mm. The phase mask filter has been applied five times. Sequence parameters are: TR = 80 ms, c~= 25~ TH = 2 mm, 16 partitions, FOV = 180 x 240 mm2, matrix size 384 x 512. The echo times range from 30 to 60 ms in steps of 10 ms. the vascular venous network surrounding the ventricles. This represents an example that images with standard resolution typically used in clinical magnetic resonance imaging examinations may fail to detect small anatomic structures, such as small vessels. In general, increasing spatial resolution in gradient echo imaging helps to reduce the influence of local static field inhomogeneities on geometrical image distortions and signal loss due to spin dephasing [13]. The two large black regions in the anterior part in the low-resolution scan in Fig. 4(a) are due to signal dephasing caused by the static susceptibility difference between brain tissue and air in the frontal cavity. No such signal loss is seen in the corresponding parts of the images with higher resolution. Note also that the si~_nal-to-noise ratio (SNR) in Fig. 4(c) is a factor of ,,/2 smaller than in Fig. 4(b). Nevertheless, the finest details are most clearly depicted in the highest resolution scan. Fig. 5 displays the case of a 16 year old female patient with a left frontal arteriovenous malformation who was examined 3 months after radiotherapy. The patient presented with headache and accompaniment of migraine, but had no episode of bleeding or seizure. As can be seen, the AVM is fed by multiple feeders from the left frontal and left middle cerebral artery and is drained into the deep venous system (Fig. 5(e)). Around the malformation, edematous and gliotic changes are visible on the conventional MR images, especially on

the T2-weighted and F L A I R images (Fig. 5(c and d)). Both the Ti-weighted images before and after application of a contrast agent do not reveal too many details of the lesion (Fig. 5(a and b)). Time-of-flight M R angiography (Fig. 5(e)) clearly shows the feeding vessels and the nidus, whereas the exact venous drainage cannot be observed. The M R venogram (Fig. 5(f)), however, depicts the venous parts of the nidus in more detail. In all cases in which the malformation was located in magnetically homogeneous parts of the brain, M R venography was valuable in the assessment of the drainage pattern of the lesion and the extent of the venous malformation was seen more clearly. On the other hand, the method had drawbacks in cases in which the imaged volume was close to air/tissue interfaces or the lower parts of the brain. Due to the long echo time signal cancellation was quite severe and the image quality not good enough for diagnosis. Bleeding represents another problem, because magnetic susceptibility variations due to the presence of hemosiderin will affect T* and will lead to undesired signal dephasing.

4. Discussion

Exploiting the small but finite differences in magnetic susceptibility of venous and arterial blood [14] makes it possible to visualize venous information separately

66

J.R. Reichenbach et al. / Magnetic Resonance Materials in Biology, Physics and Medicine 6 (1998) 62-69

Fig. 3. Influence of tip angle a on the projected venograms. Each minimum intensity projection corresponds to the same targeted volume of 22 m m The phase mask filter has been applied five times. Sequence parameters are: TR = 57 ms, TE = 40 ms, TH = 2 ram, 16 partitions, FOV = 180 x 240 rmn2, matrix size 384 x 512. The tip angles range from 15 to 35 ~

Fig. 4. Influence of resolution on the projected venograms. Each minimum intensity projection corresponds to the same targeted volume of 20 mm. The phase mask has been applied four times in each case. Sequence parameters are: TR = 58 ms, TE = 40 ms, c~= 25 ~ TH = 2 mm, F O V = 150 x 240 mm 2, 16 partitions. (a) matrix size 160 x 256, NEX = 1; (b) matrix size 320 x 512, N E X = 2 , (c) matrix size 640 x 1024, N E X = 8.

from arterial information using strongly T*-weighted high-resolution scans. Fig. 6 illustrates the underlying principle of the technique. Consider a voxel containing a small vein with a certain blood volume fraction that corresponds to an MR signal fraction of 2 and brain tissue with a signal fraction of 1 - 2. With increasing echo time the phase difference between venous blood and brain parenchyma increases and leads to a reduction of total signal S in that voxel due to partial volume effects [15,16[. The maximum magnitude difference between the two signal vectors is 22, if T2* relaxation is neglected. If the resulting signal S is still larger than the noise level in the image, the venous vessels can be indirectly inferred from the signal drop and will be visible. Application of the phase

mask filter serves to separate arteries from veins, because only veins will develop a phase difference with the surrounding parenct~yma and will be dispIayed dark in the phase images. For homogeneous brain regions or large venous vessels with diameters larger than the voxel size, the phase information will again help to display the anatomical features correctly, because in the first case the phase mask filter will not change the magnitude images upon multiplication, whereas in the second case those voxels containing venous blood will still develop a different phase and will be displayed dark in the phase images. Thus, multiplying the magnitude images with the phase mask filter will lead to suppression of venous signal inside the large vessel.

J.R. Reichenbach et al. / Magnetic Resonance Materials in Biology, Physics and Medicine 6 (1998) 62-69

67

Fig. 5. Patient with left frontal arteriovenous malformation 3 m o n t h s after radiotherapy. (a) Pre-contrast Tl-weighted SE image ( T R / T E / a / T H = 690/20/90/5, F O V = 240, M a = 384 x 512). (b) Post-contrast T~-weighted SE image. (c) T2-weighted image ( T R / T E / a / T H = 3890/90/90/5, F o r = 240, M a = 192 x 256). (d) T u r b o F L A I R image ( T R / T E / T H / T I = 3890/105/5/2340, FOV = 240, M a = 192 x 256). (e) M I P of the 3D time-of-flight M R A ( T R / T E / ~ / T H = 37/6.5/20/l, 64 partitions, FOV = 220, M a = 384 x 512). (f) m l P of the 3D, long T E sequence o'r the same volume. The phase m a s k filter was applied two times (TR/TE/c~/TH = 58/40/25/2, F O V = 240, 32 partitions, M a = 320 x 512).

As can be seen in Fig. 4(c), MR imaging has the potential to rival the superb resolution known from conventional DSA studies. Although the long acquisition time in this particular case is definitely inhibitory

for patient studies, the image indicates that MR may be able to compete with DSA in the future without the required application of a contrast agent. Future improvements in rf coil design, data processing methods,

68

J.R. Reichenbach et al./ Magnetic Resonance Materials in Biology, Physics and Medicine 6 (1998) 62 69

Fig. 6. Schematics illustrating the effect of echo time on the resulting signal vector in a single voxel which contains both blood and brain tissue. This two compartment model can be characterizedby a MR signal fraction 2 which is proportional to the blood volume fraction within the voxel and the remaining signal fraction (1 - 2) of the surrounding parenchyma. Due to the small, but finite susceptibility differencebetween venous blood and tissue, the resulting amplitude of the MR signal vector S will change as a function of echo time. For simplicity, relaxation effectshave been neglected. and increases in field strength may help to shorten the acquisition times due to improvements in signal-tonoise ratios without sacrificing the high spatial resolution, which is necessary to resolve such small structures. One further possibility is the application of a contrast agent to improve the ability to see small veins. This may simultaneously allow for shorter TE values and, thus, shorter acquisition times. However, this concept has not been investigated in the present study and still has to be tested. On the other hand, the corresponding scan leading to Fig. 4(b) with an in-plane resolution of 0.5 x 0.5 mm took less than 10 rain, which is still a tolerable acquisition time even in the case of patient studies. We feel that in most clinical cases this resolution will be sufficient to depict the venous part of lesions in great detail for diagnostic assessment. We have varied different sequence parameters in a systematic way in order to determine a set of values which give reliable and reproducible results. Careful adjustment of the sequence parameters makes it possible to acquire useful clinical data sets in less than 10 min which is still a tolerable time in most cases. The dependencies of the final venograms on parameters such as echo time, flip angle or number of phase mask multiplications, allow additional flexibility for the radiologist to optimize and tailor the information content of the final images by choosing the appropriate combination of sequence parameters. Application of the method, however, may be limited to brain regions not affected by large magnetic susceptibility inhomogenei-

ties due to the long echo time. The method is also vulnerable to motion artifacts due to the long required acquisition time. Nevertheless, no special techniques, such as ECG triggering, were applied in our volunteer studies. Only common restraining devices, such as foam pads and cushions were used. However, only experienced volunteers participated in the high resolution study (matrix 1024). They were highly motivated and did their best to keep as still as possible throughout the scans. Furthermore, for periodic or nearly periodic motion the physiological artifacts are reduced due to the effect of pseudo-gating, which occurs if the repetition time TR is much shorter than the period of the motion and the data are averaged sufficiently to eliminate the periodicity [17]. The larger the number of averages N and the smaller TR, the better the image when the motion is slightly aperiodic [17]. This condition seems to be fulfilled in the present study as demonstrated in Fig. 4(c). We have investigated several patients with arteriovenous malformations before and after radiotherapy with the new method. The amount and arrangement of venous vessels within the malformation gives important information about the AVM hemodynamics (i.e. slow or fast flow) which is of importance for the assessment of the bleeding risk. AVMs represent the most common form of vascular malformations. They consist of an intimate topographic admixture of arteries and veins with an abnormal connection between them. There are often no symptoms until complications occur and pa-

J.R. Reichenbach et a l . / Magnetic Resonance Materials in Biology, Physics and Medicine 6 (1998) 62-69

tients present with intracranial hemorrhage, seizure disorders or headaches and progressive neurologic deficits [18-21]. Magnetic resonance in AVMs is diagnostic in both the larger as well as the smaller sized lesions. It not only demonstrates the size of the vascular channels but also their precise locations. As has been shown, MR venography delineates clearly the venous part of the AVM nidus and the venous drainage patterns with high spatial resolution. Venous drainage patterns are also important for the radiotherapy planning as well as for the follow-up examinations [22]. Loss of venous structures after radiotherapy can cause significant changes in the hemodynamics of the malformation. Combined with dynamic MR angiography which makes it possible to characterize the hemodynamical aspects of an AVM, the MR venography may help to optimize the treatment planning and follow-up. Further studies are necessary to evaluate the full potential of this new technique in assessing venous malformations and other cerebral diseases. Promising candidates for this new method may be venous angiomas, cavernomas and capillary telangiectasias. Application of the method in patients suffering from brain tumor metastases would be highly interesting. Combining both conventional imaging and the proposed high resolution MR venography in a clinical protocol offers the possibility for the radiologist to tailor the entire study in appropriate cases to gain new and additional information about the venous system in therapy planning and monitoring.

Acknowledgements LR.R. thanks the RSNA Research and Education Fund for supporting this research through the 1997 RSNA Seed Grant.

References [1] Prince MR. Gadolinimn-enhanced MR aortography. Radiology 1994;191:155-64. [2] Padayachee TS, Bingham JB, Graves M J, Colchester ACF, Cox TCS. Dural sinus thrombosis. Diagnosis and follow-up by magnetic resonance angiography and imaging. Neuroradiology 1991;33:165-7. [3] Medlock MD, Olivero WC, Hanigan WC, Wright RM, Winek SJ. Children with cerebral venous thrombosis diagnosed with MRI and magnetic resonance angiography. Neurosurgery 1992;31:870-6. [4] Augustyn G, Scott J, Olson E, Gilmore R, Edwards M. Cerebral

69

venous angiomas: MR imaging. Radiology 1985;156:391-5. [5] Knopp EA. Venous Disease and tumors. MRI Clinics of North America 1995;3:509-28. [6] Stevenson J, Knopp EA, Litt AW. MP RAGE Subtraction venography: A new technique. Journal of Magnetic Resonance hnaging 1995;5:239-41. [7] Kuppusamy K, Lin W, Cizek GR, Haacke EM. In vivo regional cerebral blood volume: quantitative assessment with 3D T 1weighted pre- and postcontrast MR Imaging. Radiology 1996;201(1):106-12. [8] Wright GA, Nishimura DG, Macovski A. Flow-independent magnetic resonance projection angiography. Magn Reson Med 1991;17:126 40. [9] Cho ZH, Ro YM, Lira TH. NMR venography using the susceptibility effect produced by deoxyhemoglobin. Magn Reson Med 1992;28:25 38. [10] Ogawa S, Lee TM, Nayak AS, Glynn P. Oxygenation-sensitive contrast in magnetic resonance image of rodent brain at high magnetic fields. Magn Reson Med 1990;14:68-78. [11] Reichenbach JR, Venkatesan R, Schillinger DJ, Kido DK, Haacke EM. Small Vessels in the Human Brain: MR Venography with Deoxyhemoglobin as an Intrinsic Contrast Agent. Radiology 1997;204:272 7. [12] Salamon G, Huang YP. Radiologic Anatomy of the Brain. Berlin: Springer-Verlag, 1976. [13] Reichenbach JR, Venkatesan R, Yablonskiy DA, Thompson MR, Lai S, Haacke EM. Theory and application of static field inhomogeneity effects in gradient-echo imaging. J Magn Reson lmag 1997;7:266-79. [14] Thulborn KR, Waterton JC, Matthews PM, Radda GK. Oxygenation dependence of the transverse relaxation time of water protons in whole blood at high field. Biocbem Biophys Acta 1982;714:265-70. [15] Haacke EM, Hopkins AL, Lai S, et al. 2D and 3D high resolution gradient echo functional imaging of the brain: Venous contributions to signal in motor cortex studies. NMR Biomed 1994;7:54 62. [16] Haacke EM, Lai S, Yablonskiy DA, Lin W. In vivo validation of the BOLD mechanism: A review of signal changes in gradient echo functional MRI in the presence of flow. Int J Imag Sys Techn 1995;6(2/3):153 63. [17] Haacke EM, Lenz GW, Nelson AD. Pseudo-gating: Elimination of periodic motion artifacts in magnetic resonance imaging without gating. Magn Reson Med 1987;4:162-74. [18] McCormick WF. Pathology of vascular malformations of the brain. In: Wilson CB, Stein BM, editors. Intracranial Arteriovenous Malformations. Baltimore: Williams and Wilkins, 1984:44-63. [19] Luessenhop AJ. Natural history of cerebral arteriovenous malformations. In: Wilson CB, Stein BM, editors, lntracranial Arteriovenous Malformations. Baltimore: Williams and Wilkins, 1984:12-23. [20] Davis CH, Symon L. The management of cerebral arteriovenous malformations. Acta Neurochir 1985;74:4-11. [21] Ondra SL, Troupp H, George ED, Schwab K. The natural history of symptomatic arteriovenous malformations of the brain: A 24-year-up assessment. J Neurosurg 1990;73:387391. [22] Essig M, Engenhart R, Knopp MV, et al. Cerebral arteriovenous malformations: Improved nidus demarcation by means of dynamic tagging MR-angiography. Magn Resort Imag 1996;14:227 33~