- Email: [email protected]
Cranial MRI of small rodents using a clinical MR scanner
Methods 43 (2007) 2–11 www.elsevier.com/locate/ymeth
Cranial MRI of small rodents using a clinical MR scanner Jennifer Linn a
a,*
, Friederike Schwarz a, Christian Schichor b, Martin Wiesmann
a
Department of Neuroradiology, University Hospital Munich, Grosshadern, Marchioninistrasse 15, 81377 Munich, Germany b Department of Neurosurgery, University Hospital Munich, Grosshadern, Munich, Germany Accepted 27 March 2007
Abstract Increasing numbers of small animal models are in use in the field of neuroscience research. Magnetic resonance imaging (MRI) provides an excellent method for non-invasive imaging of the brain. Using three-dimensional (3D) MR sequences allows lesion volumetry, e.g. for the quantification of tumor size. Specialized small-bore animal MRI scanners are available for high-resolution MRI of small rodents’ brain, but major drawbacks of this dedicated equipment are its high costs and thus its limited availability. Therefore, more and more research groups use clinical MR scanners for imaging small animal models. But to achieve a reasonable spatial resolution at an acceptable signal-to-noise ratio with these scanners, some requirements concerning sequence parameters have to be matched. Thus, the aim of this paper was to present in detail a method how to perform MRI of small rodents brain using a standard clinical 1.5 T scanner and clinically available radio frequency coils to keep material costs low and to circumvent the development of custom-made coils. 2007 Elsevier Inc. All rights reserved. Keywords: Rodent models; Small animal models; Imaging; MRI; Clinical scanner
1. Introduction In the field of neuroscience, increasing numbers of small animal models are in use, for example in tumor [1–3] or stroke research as well as in the study of neurodegenerative diseases like Huntington’s or Parkinson’s disease [4], and in the study of neuroinfectious diseases [5]. Therefore, it is essential that non-invasive imaging methods are widely available to visualize the small rodent brain in vivo, for example to evaluate the extent of brain lesions, or to monitor response to therapy. Magnetic resonance imaging (MRI) provides a non-invasive imaging method offering a high spatial resolution and a high soft tissue contrast. Furthermore, it allows three-dimensional (3D) imaging e.g. for lesion volumetry. Highly specialized animal MR scanners with small bores and high magnetic field strength are available, and are in use for imaging rats and mice in vivo [6,7]. *
Corresponding author. Fax: +49 089 7095 2509. E-mail address: [email protected] (J. Linn).
1046-2023/$ - see front matter 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.ymeth.2007.03.008
The major drawback of this purpose-dedicated equipment is its high costs. Furthermore, it requires extra laboratory space, additional know-how, and personnel. As more and more institutes perform animal studies, there is an additional demand for non-invasive imaging, but not every institution can afford small animal scanners. Thus, to overcome the drawbacks of these dedicated scanners, more and more laboratories use clinical MRI scanners for imaging rats [1,8] and mice [2,3,9–12]. Clinical MRI scanners can be found as standard equipment in most hospitals and have been shown to serve as a relatively inexpensive alternative in the study of small rodent models. For imaging of small rodents with clinical MRI scanners standard head, extremity or wrist radio frequency (RF)coils [13–15] have been used. Some studies have been carried out on clinical MR scanners using custom made [1,2,12,16,17] or specially developed, commercially available RF-coils [18,19]. The aim of this paper is to present in detail a method how to perform MRI of small rodents brain using a standard clinical 1.5 T scanner and clinically available
J. Linn et al. / Methods 43 (2007) 2–11
RF-coils, and thus keeping material costs low and circumvent the development of custom-made coils. 2. Theoretical considerations MR imaging of small animal brains is technically challenging. The image quality depends mainly on signal-tonoise ratio (SNR), spatial resolution, and tissue contrast. Thus, to allow the use of clinical MRI scanners for imaging small rodents, some requirements have to be matched. First of all, SNR depends on the magnetic field strength of the MRI scanner. The higher the static magnetic field of the MR scanner, the better the SNR. Thus, the clinically widely used 1.5 and 3 T MRI scanners naturally perform at a reduced signal-to-noise ratio compared to small bore, high field (4.7 T and higher), dedicated animal scanners. Furthermore, the SNR depends on the RF-coil which is used. MR imaging of small animals can be performed using either helical coils (as are clinically used for example for the study of the human brain or knee), or small surface coils. Two different types of surface coils are feasible: small flexible coils (as are clinically used e.g. for joints), or small loop coils (e.g. clinically used for fingers). The strength of the RF signal falls off inversely with the square of the distance from the signal source. Thus, SNR can be enhanced by reducing the distance between the measured object and the receiving coil. If one uses a helical coil, this can be reached by reducing the diameter of the receiving coil in such a way that the coil is nearly completely filled by the body or head of the animal. The helical coil always should be positioned perpendicular to the magnetic field lines to optimize the signal. Surface coils allow positioning of the coil immediately above and in direct contact with the region of interest and thus are optimal for high-resolution imaging of a limited region (Fig. 1). MR images are composed of three-dimensional units, termed voxels. The geometrical dimensions of these voxels in x and y direction (in-plane resolution) and z direction
3
(slice thickness) define the spatial resolution of the image. The smaller the voxel, the higher is the spatial resolution of a certain image. Reduction of voxel size can be achieved in axial plane by reducing the field of view (FOV), and in the z dimension by reducing the slice thickness, (provided that the pixel matrix is not changed). However, there is a linear relationship between the geometrical volume of a voxel and its MR signal. Thus, if the dimensions of a voxel of 4 · 4 · 4 mm are reduced to 2 · 2 · 2 mm to increase the spatial resolution, the MR signal is reduced by a factor of 8. Thus, the drawback of reduced voxel size is that it is accompanied by a diminuation of SNR. As a consequence, the voxel size cannot be reduced to an infinite small volume. Signal averaging techniques over multiple RF excitations (=increasing the number of acquisitions) allow to overcome this problem partially. But this strategy of multiplying the RF excitations has its own practical limitations, as it results in a directly proportional prolongation of the scan time. Furthermore, the relationship between increasing the numbers of acquisition and the resulting increase in SNR is non-linear. A 2-fold increase in scan time only results in a 1.4-fold increase in SNR. Thus, the degree of image quality improvement per acquisition decreases and increasing the numbers of acquisitions too far results in a significant increase in scan time with little improvement in image quality. 2.1. Parallel imaging Another possibility to improve SNR or to decrease scanning time is the application of parallel imaging. Parallel imaging is a new technique to accelerate MRI without affecting the image contrast. This acceleration is not achieved by faster gradient hardware, but uses the spatial information of multiple receiver coils. This allows the acquisition of multiple phases encoding from a single signal echo, so that fewer phase encoding steps are needed for obtaining the same image information. This method is only
Fig. 1. Different RF-coils in relation to the animal size. (A) Standard head coil (CP Head Array Coil, Siemens Medical Solutions, Erlangen, Germany). Note the large distance between the animal’s body and the coil. (B) The double loop array coil (Double Loop Array, Siemens Medical Solutions, Erlangen, Germany) consists of a circular coil pair with an effective loop diameter of 70 mm. It requires the lower part of a standard head array coil. Both coil segments of the double loop array coil are positioned in a way that the head of the animal is fixed between the two segments. Stripes of adhesive are used for additional fixation of the two segments close to the animal’s head. (C) The small loop coil (Loop Flex Coil, small; Siemens Medical Solutions, Erlangen, Germany) has been designed for high-resolution imaging of structures near the surface. Note that it is positioned immediately above and in direct contact with the animal’s head. It is fixed with stripes of adhesive resulting in an additional fixation of the object to reduce movement artefacts induced by breathing.
4
J. Linn et al. / Methods 43 (2007) 2–11
useful if helical coils are used, and not all small helical coils can be used for parallel imaging. However, dedicated small-animal coils are available which offer parallel imaging technique and can be used with standard clinical scanners (e.g. receive-only 8-channel array surface coil, RapidBiomedical, Wu¨rzburg, Germany).
parameters. The cucumber was chosen because its configuration and size resemble the dimensions of the body of a rat, and because it offers a good signal due to its high water content.
2.2. Diffusion- and perfusion-MRI
The MRI examinations shown in this study were performed on adult male RH-RNU rats (280 g) and on adult male Wistar rats (250–300 g). The study was approved by the local ethical committee according to the local guidelines for animal care. A rat model of bacterial meningitis [25] was used to illustrate that the resolution achieved in MR imaging of small rodents on clinical scanners is sufficient to demonstrate pathologic lesions. To induce meningitis in the Wistar rats Streptococcus pneumoniae (150 ll, containing 107 colony forming units/ml) were injected transcutaneous into the cisterna magna. Control rats received 150 ml phosphate-buffered saline (PBS) intracisternally. MRI was performed 24–48 h after injection [5].
Diffusion-weighted MR imaging (DWI) and perfusionweighted MR imaging (PWI) are widely used in experimental stroke studies for non-invasive visualization of cerebral ischemia. The MR sequences used for DWI and PWI in clinical stroke imaging of the human brain (fast echo-planar (EPI) sequences) are characterized by a relatively low spatial resolution and a high sensitivity for susceptibility artifacts. These artifacts occur at the border of different tissues (e.g. between brain and bone tissue, or between skull base and air-filled sinuses). In the human brain this results in compromised image quality adjacent to the frontal or the temporal skull base. In small rodents the whole brain is so small, that it might be obscured by those artifacts, if standard clinical EPI sequences are used. However, several ways exist to minimize those susceptibility artifacts in EPI sequences: (1) First of all, great care should be taken to position the animal exactly in the centre of the scanner. This is a simple but effective measure to minimize EPI artifacts. (2) Parallel imaging allows modification of EPI sequences so that they are less prone to susceptibility artifacts but are still feasible for DWI or PWI measurements. Chen et al. demonstrated this using a parallel imaging technique (namely a generalized autocalibrating partially parallel acquisition; GRAPPA, [20]) with an acceleration factor of two [21–23]. They used a four-channel phased array wrist coil and a two-dimensional spin– echo echo-planar sequence with 12 acquisitions for DWI, and a T2*-weighted echo-planar imaging sequence with 60 measurements for PWI in combination with GRAPPA. This technique allowed the demonstration of focal cerebral ischemic lesions in DWI and PWI in rats with an acceptable amount of susceptibility artifacts and a reasonable scan time. (3) Another possibility to overcome the problem of susceptibility artifacts in DWI is to perform DWI without using EPI sequences. Spin–echo sequences with a diffusion pulse [24] can be used for this purpose as they are much less prone to artifacts. This technique has initially been used in stroke imaging in humans. The limitation of this method, however, is that those sequences require a much longer scan time as compared to EPI sequences (about 10 min versus 30 s in EPI sequences). 3. Methods 3.1. Phantom studies Phantom studies were performed using a cucumber as test object to compare different RF coils and sequence
3.2. Animals
3.3. Anaesthesia RH-RNU rats were anesthetized with a combination of Ketamine (100 mg/kg) and Xylazine (10 mg/kg ip; Rompun, Bayer AG, Leverkusen, Germany), diluted in sterile saline. The Wistar rats were anesthetized with chloral hydrate (300 mg/kg). The drugs were injected intraperitoneally (i.p.) by use of a small intravenous catheter. In order not to harm the animal, 30 G intravenous catheters (at least 22 G) are recommended. For a skilled and trained person the administration of the intraperitoneal anesthesia takes about 5 min and allows about 20– 60 min scan time. 3.4. Magnetic resonance imaging (MRI) 3.4.1. MRI scanner MRI was performed on a 1.5 T clinical MRI scanner (Magnetom Symphony, Siemens Medical solutions, Erlangen, Germany) with a field gradient strength of 30 mT/m. 3.4.2. Radio-frequency coils The aim of this paper was to demonstrate how to perform imaging of small rodents while keeping material costs low and avoiding the development of custom made coils. Thus we used two different commercially available coils: a double loop array coil (Double Loop Array, Siemens Medical Solutions, Erlangen, Germany) and a small loop coil (Loop Flex Coil, small; Siemens Medical Solutions, Erlangen, Germany). The double loop array coil consists of a circular coil pair with an effective loop diameter of 70 mm (Fig. 1). This coil is optimized for small field-ofview (FOV) imaging of the mandibular joint, wrist and eyes. It requires the lower part of a standard head array coil (CP Head Array Coil, Siemens Medical Solutions,
J. Linn et al. / Methods 43 (2007) 2–11
Erlangen, Germany) for positioning, and allows simultaneous imaging with the head coil. The small loop flex coil has been designed for high-resolution imaging of structures near the surface (e.g. fingers, toes or wrist). It weighs 100 g and has a diameter of 40 mm (Fig. 1). 3.4.3. Positioning of the animal within the scanner Anesthetized rats were placed into the centre of the magnet. To achieve best positioning of the measured object in the scanner, phantoms or animals, respectively, were mounted on a plastic holder. For imaging with the double loop array coil, which was fixed in the head coil, objects or animals were mounted on a plastic holder to position them exactly in the centre of the bore of the head coil. Both coil segments of the double-loop array coil were positioned in a way that the object or the head of the animal were fixed between the two segments (Fig. 1). If the small loop flex coil was used the object or the animal were mounted on a plastic holder with a hutch as wide as the object or animal. The small loop coil was positioned immediately above the object or the head of the animal, respectively. It was fixed with stripes of adhesive, which led to an additional fixation of the object to reduce movement artefacts induced by breathing (Fig. 1).
5
3.5. MRI sequence parameters Coronal and sagittal T1- and T2-weighted images were obtained using spin echo techniques. Furthermore 3DCISS (=constructive interference in steady state), as well as 3D-MP-RAGE (=magnetization-prepared rapid gradient-echo) sequences were performed. FOV, number of acquisitions, matrix, and slice thickness were changed systematically to illustrate their influence on SNR, resolution, and contrast. 3.6. Injection of contrast agent Contrast agent is usually injected intravenously [2,26] by direct puncture of the tail veins or by catheterization of the femoral vein. In rats the catheterization of these veins is relatively easy to perform for a trained person and ensures a rapid uptake of contrast medium. Thus, we cannulated the left femoral vein of the Wistar rats for i.v. application of the contrast agent (gadolinium–DTPA, 0.3 mmol/kg). For catheterization of the femoral vein the rat is placed in supine position and the inguinal region is prepared in a sterile fashion. After vertical incision into the skin of the groin region the subcutaneous tissue has to be removed to identify and expose the femoral vein. After that, the vein has to be ligated distally. Proximately, it is looped by a silk
Fig. 2. Results of the phantom studies. T2-weighted images performed with the double loop array coil (A) and the small loop coil (B,C), respectively. In (A) and (B) exactly the same sequence parameters were used (TR/TE = 3800/75 ms, slice-thickness = 1.5 mm, distance factor = 0.1, FOV = 60 mm, 384 · 384 pixel matrix, five acquisitions), while in (C) a smaller FOV of 60 mm compared to 100 mm in (A) and (B) was used. Note that the small loop coil (B) depicts the thin plicae (arrows) of the cucumber more precisely than the double loop array coil (A) and thus yielded better results regarding spatial resolution and SNR compared to the double loop array coil. The use of a smaller FOV (C) further enhances spatial resolution compared to B (FOV = 100 mm) and clearly demonstrates the superfine inner structure of the object (stars).
Fig. 3. Influence of the number of acquisitions and the size of the FOV on image quality. Coronal T2-weighted images (TR/TE = 3800/75 ms, slicethickness = 1.5 mm, distance factor of 0.1, 384 · 384 pixel matrix) illustrating the effect of the number of RF-excitations and the size of the FOV on image quality. (A) Five RF excitations were employed and summed for signal averaging to increase the SNR (FOV = 60 mm). (B) Only one RF excitation was performed, resulting in a reduction of the SNR (FOV = 60 mm). (C) The same sequence as shown in (A) was used, but with a FOV of 100 mm (five RFexcitations). Note the lower spatial resolution compared to (A).
6
Table 1 Influence of FOV, slice thickness, matrix and number of RF-excitations on spatial resolution, scan time and SNR FOV (mm)
Slice thickness (mm)
T1 T1 T1 T1
100 100 140 60
2 2 2 1.3
T2 T2 T2 T2 T2 T2
100 100 100 100 100 100
2 2 2 2 2 2
CISS CISS CISS CISS CISS
180 150 180 150 150
MP-RAGE MP-RAGE MP-RAGE MP-RAGE MP-RAGE
256 200 200 200 200
Distance factor (%)
Number of slices
TR/TE (ms)
Voxel size (mm3)
Matrix (pixel)
Number of RF-excitations
Scan time (min:s)
Relative SNR
10 10 10 10
14 14 14 10
462/20 462/20 462/20 462/20
0.53 · 0.52 · 2 0.78 · 0.78 · 2 1.09 · 1.09 · 2 0.16 · 0.16 · 1.3
192 · 192 128 · 128 128 · 128 384 · 384
4 4 4 10
5:49 3:54 3:54 25:40
1 1.84 1.96 0.26
0 0 0 0 0 0
14 14 14 14 14 14
3800/75 3800/75 3800/75 3800/75 3800/75 3800/75
0.31 · 0.31 · 2 0.52 · 0.52 · 2 0.78 · 0.78 · 2 0.31 · 0.31 · 2 0.31 · 0.31 · 2 0.31 · 0.31 · 2
320 · 320 192 · 192 128 · 128 320 · 320 320 · 320 320 · 320
4 4 4 3 2 1
6:53 3:54 2:50 5:10 3:28 1:46
1 2.08 3.99 0.87 0.71 0.50
0.66 0.66 05 0.5 0.5
20 20 20 20 20
80* 80* 80* 80* 80*
11.62/5.81 12.64/6.32 12.64/6.32 12.64/6.32 12.64/6.32
0.7 · 0.4 · 0.7 0.6 · 0.3 · 0.7 0.7 · 0.4 · 0.5 0.6 · 0.3 · 0.5 0.6 · 0.3 · 0.5
512 · 512 512 · 512 512 · 512 512 · 512 512 · 512
2 2 2 2 4
9 :33 9:33 9:33 10:23 20:44
1 0.69 0.76 0.53 0.74
1.5 1.5 0.8 0.8 0.8
50 50 50 50 50
120* 120* 120* 120* 120*
1580/5.0 1580/5.0 1580/5.0 1580/5.0 1580/5.0
1 · 1 · 1.5 0.8 · 0.8 · 1.5 0.8 · 0.8 · 0.8 0.8 · 0.8 · 0.8 0.8 · 0.8 · 0.8
256 · 256 256 · 256 256 · 256 256 · 256 256 · 256
1 1 1 2 4
6:46 6:46 6:46 13:30 26:59
1 0.61 0.33 0.46 0.65
FOV, field of view; RF, radio-frequency; SNR, signal-to-noise ratio; CISS, constructive interference in the steady state; MP-RAGE, magnetization-prepared rapid gradient-echo. * =slices per slab.
J. Linn et al. / Methods 43 (2007) 2–11
Sequence
J. Linn et al. / Methods 43 (2007) 2–11
tie in order to mobilize it and in order to achieve stasis of blood. Then, a longitudinal incision of a few millimeters is made into the anterior wall of the vein with vein scissors. A small polyethylene tube (PE 50, outer diameter 0.52 mm) is placed in the incision and advanced to the site of the proximal, looped silk tie. After opening the loop, the catheter can be advanced further into the vein, and the proximal silk tie is used to secure the catheter to the vein. Then the distal ligation is removed. An extension line pre-filled with contrast agent allows the injection of the gadolinium–DTPA without removing the animal from the scanner, thus avoiding repositioning. If venous catheterization is to be avoided or is more difficult to perform (e.g. in mice), the contrast agent can be injected intraperitoneally (i.p.) via a small intravenous cannula (22 G or smaller) placed i.p. (0.4–0.5 ml, 0.25 mmol/ ml; e.g. Magnevist, Schering, [3,27]). A third strategy to apply the contrast agent is subcutaneous injection (1.0 ml, 0.5 mmol/ml, [1], e.g. Magnevist, Schering). The uptake of the contrast medium takes up to 10 min using i.p. or subcutaneous injection.
7
4. Results 4.1. Coils The small loop flex coil yielded better results regarding spatial resolution and SNR compared to the double loop array coil (Fig. 2A and B). 4.2. MR sequence parameters Figs. 2 and 3 illustrate the results of the phantom and animal studies, respectively, using different number of excitations and different FOV dimensions. The sequence parameters used for T1- and T2-weighted measurements and the corresponding relative SNR are shown in Table 1. For T1-weighted imaging the highest spatial resolution at a good signal-to-noise ratio and at an acceptable scanning time, was achieved using the following imaging parameters: TR/TE = 462/20 ms, a slice-thickness of 1.3 mm, a distance factor of 0.1, a 60-mm FOV, and a 384 · 384-pixel matrix (Fig. 4A–L, coronal T1-weighted
Fig. 4. High-resolution T1-weighted images. Coronal T1-weighted images with a high spatial resolution (0.16 · 0.16 · 1.3 mm) at a good SNR. The following sequence parameters were used: TR/TE = 462/20 ms, a slice-thickness of 1.3 mm, a distance factor of 0.1, a 60-mm FOV, a 384 · 384 pixel matrix, and n = 12 slices. Ten RF excitations were employed and summed for signal averaging to increase the SNR. The acquisition time was 29:38 min.
8
J. Linn et al. / Methods 43 (2007) 2–11
images). Ten RF excitations were employed and summed for signal averaging to increase the SNR. Using these sequence parameters a spatial resolution of 0.16 · 0.16 · 1.3 mm was achieved. The acquisition time was 29:38 min. For T2-weighted images the highest spatial resolution at a good signal-to-noise ratio and at an acceptable scanning time, was achieved using the following imaging parameters: TR/TE = 3800/75 ms, a slice-thickness of 1.5 mm, a distance factor of 0.1, a 60-mm FOV, and a 384 · 384-pixel matrix (Fig. 5A–L, coronal T2-weighted images). Five RF-excitations were employed to increase the SNR. A spatial resolution of 0.16 · 0.16 · 1.5 mm (thus, an in plane resolution of 160 lm) was achieved at a scan time of 17:29 min. The above mentioned sequence parameters are recommended for T1- and T2-weighted sequences if the specific question addressed in a study requires high spatial resolu-
tion. In other studies on small rodent models at clinical MR scanners, a similar spatial resolution is reported [3,4], while dedicated small animal scanners offer resolutions between 50 and 100 lm with an adequate SNR. To reduce scan time (i.e. in glioma models were T2- and T1-weighted images pre- and post-administration of contrast are required) we propose the following modified protocol, which allows faster imaging with lower but acceptable spatial resolution: a coronal T2-weighted fast spin–echo sequence (TR/TE = 3168/96 ms, slice-thickness = 2 mm, FOV = 100 mm, 224 · 256 pixel matrix, one acquisition; scan time = 8:37 min), and a coronal T1weighted spin–echo sequence (TR/TE = 480/14 ms, slicethickness = 2 mm, FOV = 100 mm, 224 · 256 pixel matrix, one acquisition; scan time = 7:13 min), performed pre and after the administration of the contrast agent. This protocol was used for imaging of the rat model of bacterial meningitis (Fig. 6A–D).
Fig. 5. High-resolution coronal T2-weighted images. Coronal T2-weighted images with a high spatial resolution (0.16 · 0.16 · 1.5 mm) at a good SNR. The following sequence parameters were used: TR/TE = 3800/75 ms, a slice-thickness of 1.5 mm, a distance factor of 0.1, a 60-mm FOV, a 384 · 384-pixel matrix, and n = 12 slices. Five RF- excitations were employed to increase the SNR. A spatial resolution of 0.16 · 0.16 · 1.5 mm was achieved at a scan time of 17:29 min. The following anatomical structures can be easily identified: olfactory bulb (OB), neocortex (NC), thalami (T), and cerebellum (CB) as well as lateral ventricles (LV), fourth ventricle (4V), and interpeduncular cistern (IPC).
J. Linn et al. / Methods 43 (2007) 2–11
Fig. 6. Demonstration of pathologic lesions (modified from Wiesmann et al.). A rat model of bacterial meningitis was used to illustrate that the resolution achieved in MRI of small rodents on clinical scanners is sufficient to depict pathologic lesions. (A,B) Coronal T2-weighted images of a control animal (A, 24 h after injection of phosphate buffered saline into the cisterna magna) and of an infected animal (B, 48 h after injection of Streptococcus pneumoniae in the cisterna magna). The sequence parameters are: Note the marked hydrocephalus of the infected animal (arrows depict the left lateral ventricle). (C,D) Contrast-enhanced coronal T1-weighted images of a control animal (A, 24 h after injection of phosphate-buffered saline) and of an infected animal (B, 24 h after injection of S. pneumoniae). In the infected animal a significant increase in the leptomeningeal contrast enhancement (D, arrows) compared to the normal enhancement seen in the control animal (C, arrow) is demonstrated.
4.3. Lesion volumetry For lesion volumetry, the use of a 3D-CISS or a (contrast-enhanced) 3D-MP-RAGE is recommended. The following sequence parameters of the 3D-CISS sequence yielded a spatial resolution of 0.6 · 0.3 · 0.5 mm at an acquisition time of 20:44 min: TR/TE = 12.64/6.32 ms, slice-thickness = 0.66 mm, a distance factor of 20%, FOV = 150 mm, 512 · 512 pixel matrix, four acquisitions (Fig. 7A and B). For the 3D-MP-RAGE sequence the highest spatial resolution at a good signal-to-noise ratio and at an acceptable scanning time, was achieved using the following imaging
9
parameters: TR/TE = 1580/5.0 ms, a slice-thickness of 1.5 mm, a distance factor of 50%, a 200-mm FOV, and a 256 · 256 pixel matrix (Fig. 8C). Four RF-excitations were employed to increase the SNR. A spatial resolution of 0.8 · 0.8 · 0.8 mm was achieved at a scan time of 26:59 min (Fig. 7C). For post-processing the data of the 3D-sequences are transformed to a workstation (e.g. Silicon Graphics, Inc., Mountain View, CA). Tumor volumetry can be performed manually by evaluating multiplanar reconstructions in strict coronal, axial, and sagittal slice orientation using reformating software (e.g. Virtuoso, Siemens, Medical Solutions, Erlangen, Germany). The total tumor volume can be calculated as the mean of tehtumor extent in coronal, axial, and sagittal planes [28]. Furthermore, post-processing software packages are available for semi-automated 3D-volumetry (e.g. 3D Live wire software; for further information and download see webrum.uni-mannheim.de/math/skoenig/uni/download. html). The lesion has to be identified on sequential slices either by semi-automated segmentation algorithms or by encircling the lesion manually on the respective slices. Once the lesion has been identified its total volume can be estimated using simple geometrical algorithms. 4.4. Assessment of different anatomic structures of the rat brain T2-weighted images typically offer a higher contrast between grey and white matter structures, compared to T1-weighted imaging and allow identification of different anatomic structures and tissues in the rat brain. The following anatomical structures could be easily identified: olfactory bulb, neocortex, hippocampus, thalami, pons, and cerebellum as well as lateral ventricles, fourth ventricle, and interpeduncular cistern (Figs. 5 and 8). The different rat brain structures were identified according to the rat brain atlas of Paxinos and Watson [29]. T1-weighted images show only minor contrast differences between grey and white matter and are more useful if the aim of a specific study is to delineate pathologic contrast enhancement of neoplastic or inflammatory lesions (Fig. 6).
Fig. 7. Three-dimensional (3D) CISS and MP-RAGE sequences. (A,B) Coronal (A) as well as axial multiplanar reconstruction of a 3D-CISS sequence with a spatial resolution of 0.6 · 0.3 · 0.5 mm and the following sequence parameters: TR/TE = 12.64/6.32 ms, a slice-thickness of 0.5 mm, a 150-mm FOV, a 512 · 512 pixel matrix. Four RF excitations were employed and summed for signal averaging to increase the SNR. The acquisition time was 20:44 min. (C) Axial multiplanar reconstruction of a 3D-MP-RAGE sequence with a spatial resolution of 0.8 · 0.8 · 0.8 mm and the following sequence parameters: TR/TE = 1580/5.0 ms, a slice-thickness of 0.8 mm, a 200-mm FOV, a 256 · 256 pixel matrix, n = 4 RF excitations, scan time = 26:59 min.
10
J. Linn et al. / Methods 43 (2007) 2–11
Fig. 8. High-resolution sagittal T2-weighted images. Sagittal T2-weighted images with the following sequence parameters: TR/TE = 3800/75 ms, a slicethickness of 1.5 mm, a distance factor of 0.1, a 60-mm FOV, a 384 · 384-pixel matrix, n = 10 slices, and five RF-excitations. The following anatomical structures can be easily identified: olfactory bulb (OB), neocortex (NC), thalami (T), hippocampus (HC), cerebellum (CB), pons (P), medulla oblongata (MO) and cervical spinal cord (CPC) as well as lateral ventricles (LV).
5. Practical considerations
phantoms in the bore of the scanner to further improve the homogeneity of the magnetic field.
5.1. Time required 5.3. Limitations Scanning time mainly depends on the number of sequences acquired and the desired image quality as the time for a single MR sequence varies significantly (between a few and 30 min). In addition to the scanning time itself, time for anaesthesia (approximately 5 min) and correct positioning of the animal within the scanner (approximately 5 min) must be added. If several animals are to be imaged, the anaesthesia of the next animal can be induced while scanning the former one to save time. The high-resolution MRI protocol presented in this study requires approximately 50 min scan time if a T1and a T2-weighted sequence are performed, and 80 min if an additional post-contrast T1-weighted sequence is necessary. The faster protocol with reduced spatial resolution takes about 23 min (for a pre- and post-contrast T1- and a T2-weighted sequence). In case of longer measurements, animals should be kept warm, e.g. by placing a glove filled with warm water next to the animal [3]. Repetition of the i.p. injection of the anaesthetic drug can be necessary if scan time exceeds 30 min. 5.2. Ways to further improve image quality To improve image quality the magnetic field should be as homogeneous as possible. The homogeneity of the magnetic field of a large bore scanner is inferior to that of a small bore scanner. Therefore, attention should be paid to optimal shimming of the magnet. As discussed above, the object of interest should always be positioned in the centre of the magnetic field. Depending on the experimental set-up it can be beneficial to place additional water
Although we demonstrated that clinical MR scanner can be used for imaging the brain of small rodent models, there are principle technical limitations compared to high-field small animal scanners. While those dedicated animal scanners offer resolutions between 50 and 100 lm, the maximum resolution which can be achieved with an acceptable SNR using a 1.5T scanner is 120 lm. 6. Conclusions In conclusion, this study describes in detail how to perform high-resolution rat brain imaging on a clinical magnetic resonance scanner using a clinically available RF coil, demonstrating that clinical whole body MR scanners are suitable for in vivo study of small animal brains. Acknowledgments We thank Maximilian Peters and Tim Wesemann for technical assistance. References [1] F. Thorsen, L. Ersland, H. Nordli, P.Ø. Enger, P.C. Huszthy, A. Lundervold, T. Standnes, R. Bjerkvig, M. Lund-Johansen, J. Neurooncol. 63 (2003) 225–231. [2] W.R. van Furth, S. Laughlin, M.D. Taylor, B. Salhia, T. Mainprize, M. Henkelman, M.D. Cusimano, C. Ackerley, J.T. Rutka, Can. J. Neurol. Sci. 30 (2003) 326–332. [3] M.-A. Brockmann, S. Ulmer, J. Leppert, R. Nadrowitz, R. Wuestenberg, I. Nolte, D. Petersen, C. Groden, A. Giese, S. Gottschalk, Brain Res. 1068 (2006) 138–142.
J. Linn et al. / Methods 43 (2007) 2–11 [4] R. Guzman, K.O. Lo¨vblad, M. Meyer, C. Spenger, G. Schroth, H.R. Widmer, J. Neurosci. Methods 97 (2000) 77–85. [5] M. Wiesmann, U. Koedel, H. Bru¨ckmann, H.W. Pfister, Neurol. Res. 24 (2002) 307–310. [6] T. Neumann Haeflin, A. Kastrup, A. de Crespigny, M.A. Yenari, T. Ringer, G.H. Sun, M.E. Moseley, Stroke 31 (2000) 1965–1972. [7] D. Hesselbarth, C. Franke, R. Hata, G. Brinker, M. Hoehn-Berlage, NMR Biomed. 11 (1998) 423–429. [8] C. Fink, F. Kiessling, M. Bock, M.P. Lichy, B. Misselwitz, P. Peschke, N.E. Fusenig, R. Grobholz, S. Delorme, Magn. Reson. Imaging 18 (2003) 59–65. [9] N.A. Bock, N.B. Konyer, R.M. Henkelman, Magn. Reson. Med. 49 (2003) 158–167. [10] E.K. Rofstad, E. Steinsland, O. Kaalhus, Y.B. Chang, B. Hovik, H. Lyng, Int. J. Radiat. Biol. 65 (1994) 387–401. [11] D.A. Tyndall, Oral Surg. Oral Med. Oral Pathol. 76 (1993) 655–660. [12] S. Xu, T.P. Gade, C. Matei, K. Zakian, A.A. Alfieri, X. Hu, E.C. Holland, S. Soghomonian, J. Tjuvajev, D. Ballon, J.A. Koutcher, Magn. Reson. Med. 49 (2003) 551–557. [13] M.A. Acara, R.J. Mazurchuk, P.A. Nickerson, R.J. Fiel, Magn. Reson. Imaging 9 (1991) 89–92. [14] J. Lohr, R.J. Mazurchuk, M.A. Acara, P.A. Nickerson, R.J. Fiel, Magn. Reson. Imaging 9 (1991) 93–100. [15] F.A. Raila, A.P. Bowles Jr., E. Perkins, A. Terrell, J. Neurooncol. 43 (1999) 11–17. [16] R.J. Fiel, J.J. Alletto, C.M. Severin, P.A. Nickerson, M.A. Acara, R.J. Pentney, J. Magn. Reson. Imaging 1 (1991) 651–656.
11
[17] R.J. Pentney, L.J. Quackenbush, Alcohol Clin. Exp. Res. 15 (1991) 1024–1030. [18] K. Martos, D. Petersen, U. Klose, H. Requardt, R. Buchholz, P. Ohneseit, M. Schabet, K. Voig, J. Neurooncol. 14 (1992) 207–211. [19] D.A. Smith, L.P. Clarke, J.A. Fiedler, F.R. Murtagh, E.A. Bonaroti, G.J. Sengstock, G.W. Arendash, Brain Res. Bull. 31 (1993) 115–120. [20] M.A. Griswold, P.M. Jakob, R.M. Heidemann, M. Nittka, V. Jellus, J. Wang, B. Kiefer, A. Haase, Magn. Reson. Med. 47 (2002) 1202– 1210. [21] F. Chen, Y. Suzuki, N. Nagai, R. Peeters, K. Coenegrachts, W. Coudyzer, G. Marchal, Y. Ni, Radiology 233 (2004) 905–911. [22] F. Chen, Y. Suzuki, N. Nagai, R. Peeters, G. Marchal, Y. Ni, J. Neurosci. Methods 141 (2005) 55–60. [23] F. Chen, Y. Suzuki, N. Nagai, R.X. Sun, W. Coudyzer, J. Yu, G. Marchal, Y. Ni, Eur. J. Radiol. (2006), [Epub ahead of print]. [24] K.O. Lovblad, P.M. Jakob, Q. Chen, A.E. Baird, G. Schlaug, S. Warach, R.R. Edelman, Am. J. Neuroradiol. 19 (1998) 201–208. [25] U. Koedel, H.W. Pfister, Neurosci. Lett. 225 (1997) 33–36. [26] A.L. Nelson, S.A. Algon, J. Munasinghe, O. Graves, L. Goumnerova, D. Burstein, S.L. Pomeroy, J.Y. Kim, J. Neurooncol. 62 (2003) 259–267. [27] L. Taillandier, L. Antunes, K.S. Angioi-Duprez, J. Neurosci. Methods 125 (2003) 147–157. [28] R.H. Goldbrunner, M. Bendszus, M. Sasaki, T. Kraemer, K.H. Plate, K. Roosen, J.C. Tonn, Neurosurgery 47 (2000) 921–929. [29] G. Paxinos, C. Watson, The Rat Brain in Stereotaxic Coordinates, 4th ed., Academic Press, San Diego, 1998.
Cranial MRI of small rodents using a clinical MR scanner Jennifer Linn a
a,*
, Friederike Schwarz a, Christian Schichor b, Martin Wiesmann
a
Department of Neuroradiology, University Hospital Munich, Grosshadern, Marchioninistrasse 15, 81377 Munich, Germany b Department of Neurosurgery, University Hospital Munich, Grosshadern, Munich, Germany Accepted 27 March 2007
Abstract Increasing numbers of small animal models are in use in the field of neuroscience research. Magnetic resonance imaging (MRI) provides an excellent method for non-invasive imaging of the brain. Using three-dimensional (3D) MR sequences allows lesion volumetry, e.g. for the quantification of tumor size. Specialized small-bore animal MRI scanners are available for high-resolution MRI of small rodents’ brain, but major drawbacks of this dedicated equipment are its high costs and thus its limited availability. Therefore, more and more research groups use clinical MR scanners for imaging small animal models. But to achieve a reasonable spatial resolution at an acceptable signal-to-noise ratio with these scanners, some requirements concerning sequence parameters have to be matched. Thus, the aim of this paper was to present in detail a method how to perform MRI of small rodents brain using a standard clinical 1.5 T scanner and clinically available radio frequency coils to keep material costs low and to circumvent the development of custom-made coils. 2007 Elsevier Inc. All rights reserved. Keywords: Rodent models; Small animal models; Imaging; MRI; Clinical scanner
1. Introduction In the field of neuroscience, increasing numbers of small animal models are in use, for example in tumor [1–3] or stroke research as well as in the study of neurodegenerative diseases like Huntington’s or Parkinson’s disease [4], and in the study of neuroinfectious diseases [5]. Therefore, it is essential that non-invasive imaging methods are widely available to visualize the small rodent brain in vivo, for example to evaluate the extent of brain lesions, or to monitor response to therapy. Magnetic resonance imaging (MRI) provides a non-invasive imaging method offering a high spatial resolution and a high soft tissue contrast. Furthermore, it allows three-dimensional (3D) imaging e.g. for lesion volumetry. Highly specialized animal MR scanners with small bores and high magnetic field strength are available, and are in use for imaging rats and mice in vivo [6,7]. *
Corresponding author. Fax: +49 089 7095 2509. E-mail address: [email protected] (J. Linn).
1046-2023/$ - see front matter 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.ymeth.2007.03.008
The major drawback of this purpose-dedicated equipment is its high costs. Furthermore, it requires extra laboratory space, additional know-how, and personnel. As more and more institutes perform animal studies, there is an additional demand for non-invasive imaging, but not every institution can afford small animal scanners. Thus, to overcome the drawbacks of these dedicated scanners, more and more laboratories use clinical MRI scanners for imaging rats [1,8] and mice [2,3,9–12]. Clinical MRI scanners can be found as standard equipment in most hospitals and have been shown to serve as a relatively inexpensive alternative in the study of small rodent models. For imaging of small rodents with clinical MRI scanners standard head, extremity or wrist radio frequency (RF)coils [13–15] have been used. Some studies have been carried out on clinical MR scanners using custom made [1,2,12,16,17] or specially developed, commercially available RF-coils [18,19]. The aim of this paper is to present in detail a method how to perform MRI of small rodents brain using a standard clinical 1.5 T scanner and clinically available
J. Linn et al. / Methods 43 (2007) 2–11
RF-coils, and thus keeping material costs low and circumvent the development of custom-made coils. 2. Theoretical considerations MR imaging of small animal brains is technically challenging. The image quality depends mainly on signal-tonoise ratio (SNR), spatial resolution, and tissue contrast. Thus, to allow the use of clinical MRI scanners for imaging small rodents, some requirements have to be matched. First of all, SNR depends on the magnetic field strength of the MRI scanner. The higher the static magnetic field of the MR scanner, the better the SNR. Thus, the clinically widely used 1.5 and 3 T MRI scanners naturally perform at a reduced signal-to-noise ratio compared to small bore, high field (4.7 T and higher), dedicated animal scanners. Furthermore, the SNR depends on the RF-coil which is used. MR imaging of small animals can be performed using either helical coils (as are clinically used for example for the study of the human brain or knee), or small surface coils. Two different types of surface coils are feasible: small flexible coils (as are clinically used e.g. for joints), or small loop coils (e.g. clinically used for fingers). The strength of the RF signal falls off inversely with the square of the distance from the signal source. Thus, SNR can be enhanced by reducing the distance between the measured object and the receiving coil. If one uses a helical coil, this can be reached by reducing the diameter of the receiving coil in such a way that the coil is nearly completely filled by the body or head of the animal. The helical coil always should be positioned perpendicular to the magnetic field lines to optimize the signal. Surface coils allow positioning of the coil immediately above and in direct contact with the region of interest and thus are optimal for high-resolution imaging of a limited region (Fig. 1). MR images are composed of three-dimensional units, termed voxels. The geometrical dimensions of these voxels in x and y direction (in-plane resolution) and z direction
3
(slice thickness) define the spatial resolution of the image. The smaller the voxel, the higher is the spatial resolution of a certain image. Reduction of voxel size can be achieved in axial plane by reducing the field of view (FOV), and in the z dimension by reducing the slice thickness, (provided that the pixel matrix is not changed). However, there is a linear relationship between the geometrical volume of a voxel and its MR signal. Thus, if the dimensions of a voxel of 4 · 4 · 4 mm are reduced to 2 · 2 · 2 mm to increase the spatial resolution, the MR signal is reduced by a factor of 8. Thus, the drawback of reduced voxel size is that it is accompanied by a diminuation of SNR. As a consequence, the voxel size cannot be reduced to an infinite small volume. Signal averaging techniques over multiple RF excitations (=increasing the number of acquisitions) allow to overcome this problem partially. But this strategy of multiplying the RF excitations has its own practical limitations, as it results in a directly proportional prolongation of the scan time. Furthermore, the relationship between increasing the numbers of acquisition and the resulting increase in SNR is non-linear. A 2-fold increase in scan time only results in a 1.4-fold increase in SNR. Thus, the degree of image quality improvement per acquisition decreases and increasing the numbers of acquisitions too far results in a significant increase in scan time with little improvement in image quality. 2.1. Parallel imaging Another possibility to improve SNR or to decrease scanning time is the application of parallel imaging. Parallel imaging is a new technique to accelerate MRI without affecting the image contrast. This acceleration is not achieved by faster gradient hardware, but uses the spatial information of multiple receiver coils. This allows the acquisition of multiple phases encoding from a single signal echo, so that fewer phase encoding steps are needed for obtaining the same image information. This method is only
Fig. 1. Different RF-coils in relation to the animal size. (A) Standard head coil (CP Head Array Coil, Siemens Medical Solutions, Erlangen, Germany). Note the large distance between the animal’s body and the coil. (B) The double loop array coil (Double Loop Array, Siemens Medical Solutions, Erlangen, Germany) consists of a circular coil pair with an effective loop diameter of 70 mm. It requires the lower part of a standard head array coil. Both coil segments of the double loop array coil are positioned in a way that the head of the animal is fixed between the two segments. Stripes of adhesive are used for additional fixation of the two segments close to the animal’s head. (C) The small loop coil (Loop Flex Coil, small; Siemens Medical Solutions, Erlangen, Germany) has been designed for high-resolution imaging of structures near the surface. Note that it is positioned immediately above and in direct contact with the animal’s head. It is fixed with stripes of adhesive resulting in an additional fixation of the object to reduce movement artefacts induced by breathing.
4
J. Linn et al. / Methods 43 (2007) 2–11
useful if helical coils are used, and not all small helical coils can be used for parallel imaging. However, dedicated small-animal coils are available which offer parallel imaging technique and can be used with standard clinical scanners (e.g. receive-only 8-channel array surface coil, RapidBiomedical, Wu¨rzburg, Germany).
parameters. The cucumber was chosen because its configuration and size resemble the dimensions of the body of a rat, and because it offers a good signal due to its high water content.
2.2. Diffusion- and perfusion-MRI
The MRI examinations shown in this study were performed on adult male RH-RNU rats (280 g) and on adult male Wistar rats (250–300 g). The study was approved by the local ethical committee according to the local guidelines for animal care. A rat model of bacterial meningitis [25] was used to illustrate that the resolution achieved in MR imaging of small rodents on clinical scanners is sufficient to demonstrate pathologic lesions. To induce meningitis in the Wistar rats Streptococcus pneumoniae (150 ll, containing 107 colony forming units/ml) were injected transcutaneous into the cisterna magna. Control rats received 150 ml phosphate-buffered saline (PBS) intracisternally. MRI was performed 24–48 h after injection [5].
Diffusion-weighted MR imaging (DWI) and perfusionweighted MR imaging (PWI) are widely used in experimental stroke studies for non-invasive visualization of cerebral ischemia. The MR sequences used for DWI and PWI in clinical stroke imaging of the human brain (fast echo-planar (EPI) sequences) are characterized by a relatively low spatial resolution and a high sensitivity for susceptibility artifacts. These artifacts occur at the border of different tissues (e.g. between brain and bone tissue, or between skull base and air-filled sinuses). In the human brain this results in compromised image quality adjacent to the frontal or the temporal skull base. In small rodents the whole brain is so small, that it might be obscured by those artifacts, if standard clinical EPI sequences are used. However, several ways exist to minimize those susceptibility artifacts in EPI sequences: (1) First of all, great care should be taken to position the animal exactly in the centre of the scanner. This is a simple but effective measure to minimize EPI artifacts. (2) Parallel imaging allows modification of EPI sequences so that they are less prone to susceptibility artifacts but are still feasible for DWI or PWI measurements. Chen et al. demonstrated this using a parallel imaging technique (namely a generalized autocalibrating partially parallel acquisition; GRAPPA, [20]) with an acceleration factor of two [21–23]. They used a four-channel phased array wrist coil and a two-dimensional spin– echo echo-planar sequence with 12 acquisitions for DWI, and a T2*-weighted echo-planar imaging sequence with 60 measurements for PWI in combination with GRAPPA. This technique allowed the demonstration of focal cerebral ischemic lesions in DWI and PWI in rats with an acceptable amount of susceptibility artifacts and a reasonable scan time. (3) Another possibility to overcome the problem of susceptibility artifacts in DWI is to perform DWI without using EPI sequences. Spin–echo sequences with a diffusion pulse [24] can be used for this purpose as they are much less prone to artifacts. This technique has initially been used in stroke imaging in humans. The limitation of this method, however, is that those sequences require a much longer scan time as compared to EPI sequences (about 10 min versus 30 s in EPI sequences). 3. Methods 3.1. Phantom studies Phantom studies were performed using a cucumber as test object to compare different RF coils and sequence
3.2. Animals
3.3. Anaesthesia RH-RNU rats were anesthetized with a combination of Ketamine (100 mg/kg) and Xylazine (10 mg/kg ip; Rompun, Bayer AG, Leverkusen, Germany), diluted in sterile saline. The Wistar rats were anesthetized with chloral hydrate (300 mg/kg). The drugs were injected intraperitoneally (i.p.) by use of a small intravenous catheter. In order not to harm the animal, 30 G intravenous catheters (at least 22 G) are recommended. For a skilled and trained person the administration of the intraperitoneal anesthesia takes about 5 min and allows about 20– 60 min scan time. 3.4. Magnetic resonance imaging (MRI) 3.4.1. MRI scanner MRI was performed on a 1.5 T clinical MRI scanner (Magnetom Symphony, Siemens Medical solutions, Erlangen, Germany) with a field gradient strength of 30 mT/m. 3.4.2. Radio-frequency coils The aim of this paper was to demonstrate how to perform imaging of small rodents while keeping material costs low and avoiding the development of custom made coils. Thus we used two different commercially available coils: a double loop array coil (Double Loop Array, Siemens Medical Solutions, Erlangen, Germany) and a small loop coil (Loop Flex Coil, small; Siemens Medical Solutions, Erlangen, Germany). The double loop array coil consists of a circular coil pair with an effective loop diameter of 70 mm (Fig. 1). This coil is optimized for small field-ofview (FOV) imaging of the mandibular joint, wrist and eyes. It requires the lower part of a standard head array coil (CP Head Array Coil, Siemens Medical Solutions,
J. Linn et al. / Methods 43 (2007) 2–11
Erlangen, Germany) for positioning, and allows simultaneous imaging with the head coil. The small loop flex coil has been designed for high-resolution imaging of structures near the surface (e.g. fingers, toes or wrist). It weighs 100 g and has a diameter of 40 mm (Fig. 1). 3.4.3. Positioning of the animal within the scanner Anesthetized rats were placed into the centre of the magnet. To achieve best positioning of the measured object in the scanner, phantoms or animals, respectively, were mounted on a plastic holder. For imaging with the double loop array coil, which was fixed in the head coil, objects or animals were mounted on a plastic holder to position them exactly in the centre of the bore of the head coil. Both coil segments of the double-loop array coil were positioned in a way that the object or the head of the animal were fixed between the two segments (Fig. 1). If the small loop flex coil was used the object or the animal were mounted on a plastic holder with a hutch as wide as the object or animal. The small loop coil was positioned immediately above the object or the head of the animal, respectively. It was fixed with stripes of adhesive, which led to an additional fixation of the object to reduce movement artefacts induced by breathing (Fig. 1).
5
3.5. MRI sequence parameters Coronal and sagittal T1- and T2-weighted images were obtained using spin echo techniques. Furthermore 3DCISS (=constructive interference in steady state), as well as 3D-MP-RAGE (=magnetization-prepared rapid gradient-echo) sequences were performed. FOV, number of acquisitions, matrix, and slice thickness were changed systematically to illustrate their influence on SNR, resolution, and contrast. 3.6. Injection of contrast agent Contrast agent is usually injected intravenously [2,26] by direct puncture of the tail veins or by catheterization of the femoral vein. In rats the catheterization of these veins is relatively easy to perform for a trained person and ensures a rapid uptake of contrast medium. Thus, we cannulated the left femoral vein of the Wistar rats for i.v. application of the contrast agent (gadolinium–DTPA, 0.3 mmol/kg). For catheterization of the femoral vein the rat is placed in supine position and the inguinal region is prepared in a sterile fashion. After vertical incision into the skin of the groin region the subcutaneous tissue has to be removed to identify and expose the femoral vein. After that, the vein has to be ligated distally. Proximately, it is looped by a silk
Fig. 2. Results of the phantom studies. T2-weighted images performed with the double loop array coil (A) and the small loop coil (B,C), respectively. In (A) and (B) exactly the same sequence parameters were used (TR/TE = 3800/75 ms, slice-thickness = 1.5 mm, distance factor = 0.1, FOV = 60 mm, 384 · 384 pixel matrix, five acquisitions), while in (C) a smaller FOV of 60 mm compared to 100 mm in (A) and (B) was used. Note that the small loop coil (B) depicts the thin plicae (arrows) of the cucumber more precisely than the double loop array coil (A) and thus yielded better results regarding spatial resolution and SNR compared to the double loop array coil. The use of a smaller FOV (C) further enhances spatial resolution compared to B (FOV = 100 mm) and clearly demonstrates the superfine inner structure of the object (stars).
Fig. 3. Influence of the number of acquisitions and the size of the FOV on image quality. Coronal T2-weighted images (TR/TE = 3800/75 ms, slicethickness = 1.5 mm, distance factor of 0.1, 384 · 384 pixel matrix) illustrating the effect of the number of RF-excitations and the size of the FOV on image quality. (A) Five RF excitations were employed and summed for signal averaging to increase the SNR (FOV = 60 mm). (B) Only one RF excitation was performed, resulting in a reduction of the SNR (FOV = 60 mm). (C) The same sequence as shown in (A) was used, but with a FOV of 100 mm (five RFexcitations). Note the lower spatial resolution compared to (A).
6
Table 1 Influence of FOV, slice thickness, matrix and number of RF-excitations on spatial resolution, scan time and SNR FOV (mm)
Slice thickness (mm)
T1 T1 T1 T1
100 100 140 60
2 2 2 1.3
T2 T2 T2 T2 T2 T2
100 100 100 100 100 100
2 2 2 2 2 2
CISS CISS CISS CISS CISS
180 150 180 150 150
MP-RAGE MP-RAGE MP-RAGE MP-RAGE MP-RAGE
256 200 200 200 200
Distance factor (%)
Number of slices
TR/TE (ms)
Voxel size (mm3)
Matrix (pixel)
Number of RF-excitations
Scan time (min:s)
Relative SNR
10 10 10 10
14 14 14 10
462/20 462/20 462/20 462/20
0.53 · 0.52 · 2 0.78 · 0.78 · 2 1.09 · 1.09 · 2 0.16 · 0.16 · 1.3
192 · 192 128 · 128 128 · 128 384 · 384
4 4 4 10
5:49 3:54 3:54 25:40
1 1.84 1.96 0.26
0 0 0 0 0 0
14 14 14 14 14 14
3800/75 3800/75 3800/75 3800/75 3800/75 3800/75
0.31 · 0.31 · 2 0.52 · 0.52 · 2 0.78 · 0.78 · 2 0.31 · 0.31 · 2 0.31 · 0.31 · 2 0.31 · 0.31 · 2
320 · 320 192 · 192 128 · 128 320 · 320 320 · 320 320 · 320
4 4 4 3 2 1
6:53 3:54 2:50 5:10 3:28 1:46
1 2.08 3.99 0.87 0.71 0.50
0.66 0.66 05 0.5 0.5
20 20 20 20 20
80* 80* 80* 80* 80*
11.62/5.81 12.64/6.32 12.64/6.32 12.64/6.32 12.64/6.32
0.7 · 0.4 · 0.7 0.6 · 0.3 · 0.7 0.7 · 0.4 · 0.5 0.6 · 0.3 · 0.5 0.6 · 0.3 · 0.5
512 · 512 512 · 512 512 · 512 512 · 512 512 · 512
2 2 2 2 4
9 :33 9:33 9:33 10:23 20:44
1 0.69 0.76 0.53 0.74
1.5 1.5 0.8 0.8 0.8
50 50 50 50 50
120* 120* 120* 120* 120*
1580/5.0 1580/5.0 1580/5.0 1580/5.0 1580/5.0
1 · 1 · 1.5 0.8 · 0.8 · 1.5 0.8 · 0.8 · 0.8 0.8 · 0.8 · 0.8 0.8 · 0.8 · 0.8
256 · 256 256 · 256 256 · 256 256 · 256 256 · 256
1 1 1 2 4
6:46 6:46 6:46 13:30 26:59
1 0.61 0.33 0.46 0.65
FOV, field of view; RF, radio-frequency; SNR, signal-to-noise ratio; CISS, constructive interference in the steady state; MP-RAGE, magnetization-prepared rapid gradient-echo. * =slices per slab.
J. Linn et al. / Methods 43 (2007) 2–11
Sequence
J. Linn et al. / Methods 43 (2007) 2–11
tie in order to mobilize it and in order to achieve stasis of blood. Then, a longitudinal incision of a few millimeters is made into the anterior wall of the vein with vein scissors. A small polyethylene tube (PE 50, outer diameter 0.52 mm) is placed in the incision and advanced to the site of the proximal, looped silk tie. After opening the loop, the catheter can be advanced further into the vein, and the proximal silk tie is used to secure the catheter to the vein. Then the distal ligation is removed. An extension line pre-filled with contrast agent allows the injection of the gadolinium–DTPA without removing the animal from the scanner, thus avoiding repositioning. If venous catheterization is to be avoided or is more difficult to perform (e.g. in mice), the contrast agent can be injected intraperitoneally (i.p.) via a small intravenous cannula (22 G or smaller) placed i.p. (0.4–0.5 ml, 0.25 mmol/ ml; e.g. Magnevist, Schering, [3,27]). A third strategy to apply the contrast agent is subcutaneous injection (1.0 ml, 0.5 mmol/ml, [1], e.g. Magnevist, Schering). The uptake of the contrast medium takes up to 10 min using i.p. or subcutaneous injection.
7
4. Results 4.1. Coils The small loop flex coil yielded better results regarding spatial resolution and SNR compared to the double loop array coil (Fig. 2A and B). 4.2. MR sequence parameters Figs. 2 and 3 illustrate the results of the phantom and animal studies, respectively, using different number of excitations and different FOV dimensions. The sequence parameters used for T1- and T2-weighted measurements and the corresponding relative SNR are shown in Table 1. For T1-weighted imaging the highest spatial resolution at a good signal-to-noise ratio and at an acceptable scanning time, was achieved using the following imaging parameters: TR/TE = 462/20 ms, a slice-thickness of 1.3 mm, a distance factor of 0.1, a 60-mm FOV, and a 384 · 384-pixel matrix (Fig. 4A–L, coronal T1-weighted
Fig. 4. High-resolution T1-weighted images. Coronal T1-weighted images with a high spatial resolution (0.16 · 0.16 · 1.3 mm) at a good SNR. The following sequence parameters were used: TR/TE = 462/20 ms, a slice-thickness of 1.3 mm, a distance factor of 0.1, a 60-mm FOV, a 384 · 384 pixel matrix, and n = 12 slices. Ten RF excitations were employed and summed for signal averaging to increase the SNR. The acquisition time was 29:38 min.
8
J. Linn et al. / Methods 43 (2007) 2–11
images). Ten RF excitations were employed and summed for signal averaging to increase the SNR. Using these sequence parameters a spatial resolution of 0.16 · 0.16 · 1.3 mm was achieved. The acquisition time was 29:38 min. For T2-weighted images the highest spatial resolution at a good signal-to-noise ratio and at an acceptable scanning time, was achieved using the following imaging parameters: TR/TE = 3800/75 ms, a slice-thickness of 1.5 mm, a distance factor of 0.1, a 60-mm FOV, and a 384 · 384-pixel matrix (Fig. 5A–L, coronal T2-weighted images). Five RF-excitations were employed to increase the SNR. A spatial resolution of 0.16 · 0.16 · 1.5 mm (thus, an in plane resolution of 160 lm) was achieved at a scan time of 17:29 min. The above mentioned sequence parameters are recommended for T1- and T2-weighted sequences if the specific question addressed in a study requires high spatial resolu-
tion. In other studies on small rodent models at clinical MR scanners, a similar spatial resolution is reported [3,4], while dedicated small animal scanners offer resolutions between 50 and 100 lm with an adequate SNR. To reduce scan time (i.e. in glioma models were T2- and T1-weighted images pre- and post-administration of contrast are required) we propose the following modified protocol, which allows faster imaging with lower but acceptable spatial resolution: a coronal T2-weighted fast spin–echo sequence (TR/TE = 3168/96 ms, slice-thickness = 2 mm, FOV = 100 mm, 224 · 256 pixel matrix, one acquisition; scan time = 8:37 min), and a coronal T1weighted spin–echo sequence (TR/TE = 480/14 ms, slicethickness = 2 mm, FOV = 100 mm, 224 · 256 pixel matrix, one acquisition; scan time = 7:13 min), performed pre and after the administration of the contrast agent. This protocol was used for imaging of the rat model of bacterial meningitis (Fig. 6A–D).
Fig. 5. High-resolution coronal T2-weighted images. Coronal T2-weighted images with a high spatial resolution (0.16 · 0.16 · 1.5 mm) at a good SNR. The following sequence parameters were used: TR/TE = 3800/75 ms, a slice-thickness of 1.5 mm, a distance factor of 0.1, a 60-mm FOV, a 384 · 384-pixel matrix, and n = 12 slices. Five RF- excitations were employed to increase the SNR. A spatial resolution of 0.16 · 0.16 · 1.5 mm was achieved at a scan time of 17:29 min. The following anatomical structures can be easily identified: olfactory bulb (OB), neocortex (NC), thalami (T), and cerebellum (CB) as well as lateral ventricles (LV), fourth ventricle (4V), and interpeduncular cistern (IPC).
J. Linn et al. / Methods 43 (2007) 2–11
Fig. 6. Demonstration of pathologic lesions (modified from Wiesmann et al.). A rat model of bacterial meningitis was used to illustrate that the resolution achieved in MRI of small rodents on clinical scanners is sufficient to depict pathologic lesions. (A,B) Coronal T2-weighted images of a control animal (A, 24 h after injection of phosphate buffered saline into the cisterna magna) and of an infected animal (B, 48 h after injection of Streptococcus pneumoniae in the cisterna magna). The sequence parameters are: Note the marked hydrocephalus of the infected animal (arrows depict the left lateral ventricle). (C,D) Contrast-enhanced coronal T1-weighted images of a control animal (A, 24 h after injection of phosphate-buffered saline) and of an infected animal (B, 24 h after injection of S. pneumoniae). In the infected animal a significant increase in the leptomeningeal contrast enhancement (D, arrows) compared to the normal enhancement seen in the control animal (C, arrow) is demonstrated.
4.3. Lesion volumetry For lesion volumetry, the use of a 3D-CISS or a (contrast-enhanced) 3D-MP-RAGE is recommended. The following sequence parameters of the 3D-CISS sequence yielded a spatial resolution of 0.6 · 0.3 · 0.5 mm at an acquisition time of 20:44 min: TR/TE = 12.64/6.32 ms, slice-thickness = 0.66 mm, a distance factor of 20%, FOV = 150 mm, 512 · 512 pixel matrix, four acquisitions (Fig. 7A and B). For the 3D-MP-RAGE sequence the highest spatial resolution at a good signal-to-noise ratio and at an acceptable scanning time, was achieved using the following imaging
9
parameters: TR/TE = 1580/5.0 ms, a slice-thickness of 1.5 mm, a distance factor of 50%, a 200-mm FOV, and a 256 · 256 pixel matrix (Fig. 8C). Four RF-excitations were employed to increase the SNR. A spatial resolution of 0.8 · 0.8 · 0.8 mm was achieved at a scan time of 26:59 min (Fig. 7C). For post-processing the data of the 3D-sequences are transformed to a workstation (e.g. Silicon Graphics, Inc., Mountain View, CA). Tumor volumetry can be performed manually by evaluating multiplanar reconstructions in strict coronal, axial, and sagittal slice orientation using reformating software (e.g. Virtuoso, Siemens, Medical Solutions, Erlangen, Germany). The total tumor volume can be calculated as the mean of tehtumor extent in coronal, axial, and sagittal planes [28]. Furthermore, post-processing software packages are available for semi-automated 3D-volumetry (e.g. 3D Live wire software; for further information and download see webrum.uni-mannheim.de/math/skoenig/uni/download. html). The lesion has to be identified on sequential slices either by semi-automated segmentation algorithms or by encircling the lesion manually on the respective slices. Once the lesion has been identified its total volume can be estimated using simple geometrical algorithms. 4.4. Assessment of different anatomic structures of the rat brain T2-weighted images typically offer a higher contrast between grey and white matter structures, compared to T1-weighted imaging and allow identification of different anatomic structures and tissues in the rat brain. The following anatomical structures could be easily identified: olfactory bulb, neocortex, hippocampus, thalami, pons, and cerebellum as well as lateral ventricles, fourth ventricle, and interpeduncular cistern (Figs. 5 and 8). The different rat brain structures were identified according to the rat brain atlas of Paxinos and Watson [29]. T1-weighted images show only minor contrast differences between grey and white matter and are more useful if the aim of a specific study is to delineate pathologic contrast enhancement of neoplastic or inflammatory lesions (Fig. 6).
Fig. 7. Three-dimensional (3D) CISS and MP-RAGE sequences. (A,B) Coronal (A) as well as axial multiplanar reconstruction of a 3D-CISS sequence with a spatial resolution of 0.6 · 0.3 · 0.5 mm and the following sequence parameters: TR/TE = 12.64/6.32 ms, a slice-thickness of 0.5 mm, a 150-mm FOV, a 512 · 512 pixel matrix. Four RF excitations were employed and summed for signal averaging to increase the SNR. The acquisition time was 20:44 min. (C) Axial multiplanar reconstruction of a 3D-MP-RAGE sequence with a spatial resolution of 0.8 · 0.8 · 0.8 mm and the following sequence parameters: TR/TE = 1580/5.0 ms, a slice-thickness of 0.8 mm, a 200-mm FOV, a 256 · 256 pixel matrix, n = 4 RF excitations, scan time = 26:59 min.
10
J. Linn et al. / Methods 43 (2007) 2–11
Fig. 8. High-resolution sagittal T2-weighted images. Sagittal T2-weighted images with the following sequence parameters: TR/TE = 3800/75 ms, a slicethickness of 1.5 mm, a distance factor of 0.1, a 60-mm FOV, a 384 · 384-pixel matrix, n = 10 slices, and five RF-excitations. The following anatomical structures can be easily identified: olfactory bulb (OB), neocortex (NC), thalami (T), hippocampus (HC), cerebellum (CB), pons (P), medulla oblongata (MO) and cervical spinal cord (CPC) as well as lateral ventricles (LV).
5. Practical considerations
phantoms in the bore of the scanner to further improve the homogeneity of the magnetic field.
5.1. Time required 5.3. Limitations Scanning time mainly depends on the number of sequences acquired and the desired image quality as the time for a single MR sequence varies significantly (between a few and 30 min). In addition to the scanning time itself, time for anaesthesia (approximately 5 min) and correct positioning of the animal within the scanner (approximately 5 min) must be added. If several animals are to be imaged, the anaesthesia of the next animal can be induced while scanning the former one to save time. The high-resolution MRI protocol presented in this study requires approximately 50 min scan time if a T1and a T2-weighted sequence are performed, and 80 min if an additional post-contrast T1-weighted sequence is necessary. The faster protocol with reduced spatial resolution takes about 23 min (for a pre- and post-contrast T1- and a T2-weighted sequence). In case of longer measurements, animals should be kept warm, e.g. by placing a glove filled with warm water next to the animal [3]. Repetition of the i.p. injection of the anaesthetic drug can be necessary if scan time exceeds 30 min. 5.2. Ways to further improve image quality To improve image quality the magnetic field should be as homogeneous as possible. The homogeneity of the magnetic field of a large bore scanner is inferior to that of a small bore scanner. Therefore, attention should be paid to optimal shimming of the magnet. As discussed above, the object of interest should always be positioned in the centre of the magnetic field. Depending on the experimental set-up it can be beneficial to place additional water
Although we demonstrated that clinical MR scanner can be used for imaging the brain of small rodent models, there are principle technical limitations compared to high-field small animal scanners. While those dedicated animal scanners offer resolutions between 50 and 100 lm, the maximum resolution which can be achieved with an acceptable SNR using a 1.5T scanner is 120 lm. 6. Conclusions In conclusion, this study describes in detail how to perform high-resolution rat brain imaging on a clinical magnetic resonance scanner using a clinically available RF coil, demonstrating that clinical whole body MR scanners are suitable for in vivo study of small animal brains. Acknowledgments We thank Maximilian Peters and Tim Wesemann for technical assistance. References [1] F. Thorsen, L. Ersland, H. Nordli, P.Ø. Enger, P.C. Huszthy, A. Lundervold, T. Standnes, R. Bjerkvig, M. Lund-Johansen, J. Neurooncol. 63 (2003) 225–231. [2] W.R. van Furth, S. Laughlin, M.D. Taylor, B. Salhia, T. Mainprize, M. Henkelman, M.D. Cusimano, C. Ackerley, J.T. Rutka, Can. J. Neurol. Sci. 30 (2003) 326–332. [3] M.-A. Brockmann, S. Ulmer, J. Leppert, R. Nadrowitz, R. Wuestenberg, I. Nolte, D. Petersen, C. Groden, A. Giese, S. Gottschalk, Brain Res. 1068 (2006) 138–142.
J. Linn et al. / Methods 43 (2007) 2–11 [4] R. Guzman, K.O. Lo¨vblad, M. Meyer, C. Spenger, G. Schroth, H.R. Widmer, J. Neurosci. Methods 97 (2000) 77–85. [5] M. Wiesmann, U. Koedel, H. Bru¨ckmann, H.W. Pfister, Neurol. Res. 24 (2002) 307–310. [6] T. Neumann Haeflin, A. Kastrup, A. de Crespigny, M.A. Yenari, T. Ringer, G.H. Sun, M.E. Moseley, Stroke 31 (2000) 1965–1972. [7] D. Hesselbarth, C. Franke, R. Hata, G. Brinker, M. Hoehn-Berlage, NMR Biomed. 11 (1998) 423–429. [8] C. Fink, F. Kiessling, M. Bock, M.P. Lichy, B. Misselwitz, P. Peschke, N.E. Fusenig, R. Grobholz, S. Delorme, Magn. Reson. Imaging 18 (2003) 59–65. [9] N.A. Bock, N.B. Konyer, R.M. Henkelman, Magn. Reson. Med. 49 (2003) 158–167. [10] E.K. Rofstad, E. Steinsland, O. Kaalhus, Y.B. Chang, B. Hovik, H. Lyng, Int. J. Radiat. Biol. 65 (1994) 387–401. [11] D.A. Tyndall, Oral Surg. Oral Med. Oral Pathol. 76 (1993) 655–660. [12] S. Xu, T.P. Gade, C. Matei, K. Zakian, A.A. Alfieri, X. Hu, E.C. Holland, S. Soghomonian, J. Tjuvajev, D. Ballon, J.A. Koutcher, Magn. Reson. Med. 49 (2003) 551–557. [13] M.A. Acara, R.J. Mazurchuk, P.A. Nickerson, R.J. Fiel, Magn. Reson. Imaging 9 (1991) 89–92. [14] J. Lohr, R.J. Mazurchuk, M.A. Acara, P.A. Nickerson, R.J. Fiel, Magn. Reson. Imaging 9 (1991) 93–100. [15] F.A. Raila, A.P. Bowles Jr., E. Perkins, A. Terrell, J. Neurooncol. 43 (1999) 11–17. [16] R.J. Fiel, J.J. Alletto, C.M. Severin, P.A. Nickerson, M.A. Acara, R.J. Pentney, J. Magn. Reson. Imaging 1 (1991) 651–656.
11
[17] R.J. Pentney, L.J. Quackenbush, Alcohol Clin. Exp. Res. 15 (1991) 1024–1030. [18] K. Martos, D. Petersen, U. Klose, H. Requardt, R. Buchholz, P. Ohneseit, M. Schabet, K. Voig, J. Neurooncol. 14 (1992) 207–211. [19] D.A. Smith, L.P. Clarke, J.A. Fiedler, F.R. Murtagh, E.A. Bonaroti, G.J. Sengstock, G.W. Arendash, Brain Res. Bull. 31 (1993) 115–120. [20] M.A. Griswold, P.M. Jakob, R.M. Heidemann, M. Nittka, V. Jellus, J. Wang, B. Kiefer, A. Haase, Magn. Reson. Med. 47 (2002) 1202– 1210. [21] F. Chen, Y. Suzuki, N. Nagai, R. Peeters, K. Coenegrachts, W. Coudyzer, G. Marchal, Y. Ni, Radiology 233 (2004) 905–911. [22] F. Chen, Y. Suzuki, N. Nagai, R. Peeters, G. Marchal, Y. Ni, J. Neurosci. Methods 141 (2005) 55–60. [23] F. Chen, Y. Suzuki, N. Nagai, R.X. Sun, W. Coudyzer, J. Yu, G. Marchal, Y. Ni, Eur. J. Radiol. (2006), [Epub ahead of print]. [24] K.O. Lovblad, P.M. Jakob, Q. Chen, A.E. Baird, G. Schlaug, S. Warach, R.R. Edelman, Am. J. Neuroradiol. 19 (1998) 201–208. [25] U. Koedel, H.W. Pfister, Neurosci. Lett. 225 (1997) 33–36. [26] A.L. Nelson, S.A. Algon, J. Munasinghe, O. Graves, L. Goumnerova, D. Burstein, S.L. Pomeroy, J.Y. Kim, J. Neurooncol. 62 (2003) 259–267. [27] L. Taillandier, L. Antunes, K.S. Angioi-Duprez, J. Neurosci. Methods 125 (2003) 147–157. [28] R.H. Goldbrunner, M. Bendszus, M. Sasaki, T. Kraemer, K.H. Plate, K. Roosen, J.C. Tonn, Neurosurgery 47 (2000) 921–929. [29] G. Paxinos, C. Watson, The Rat Brain in Stereotaxic Coordinates, 4th ed., Academic Press, San Diego, 1998.