| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Technical Developments |
1 From the Dept of Radiology, Box 800170, University of Virginia School of Medicine, Charlottesville, VA 22908 (J.P.M., J.R.B.); Dept of Nuclear Engineering, Massachusetts Institute of Technology, Cambridge (S.B.); and Depts of Radiology, Children’s Hospital (R.V.M., R.L.R.) and Brigham and Women’s Hospital (R.V.M., C.R.G.G., F.A.J.), Harvard Medical School, Boston, Mass. Received Jan 4, 1999; revision requested Mar 17; final revision received Nov 26; accepted Dec 17. J.P.M., S.B., R.V.M., C.R.G.G., F.A.J., and J.R.B. supported in part by grant NS-35142 from the National Institute of Neurological Disorders and Stroke. Address correspondence to J.P.M. (e-mail: jpm7r@virginia.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMaterials and Methods Results DiscussionREFERENCES |
|---|
Index terms: Brain, abnormalities, 10.121411, 10.363, 10.3637, 10.871 • Brain, MR, 10.121411, 10.121413, 10.121416 • Magnetic resonance (MR), pulse sequences, 10.121411, 10.121413, 10.121416 • Magnetic resonance (MR), rapid imaging, 10.121411, 10.121413, 10.121416 • Magnetic resonance (MR), three-dimensional, 10.121411, 10.121413, 10.121416
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionREFERENCES |
|---|
In practice, the limitations of three-dimensional (3D) techniques have hindered their routine use for imaging the brain. The contrast of 3D gradient-recalled-echo techniques is, in many instances, unsatisfactory when compared with that of SE imaging. The 3D gradient-recalled-echo techniques are more susceptible to image artifacts from off-resonance effects, and true T2 weighting is difficult to attain. As a result, only short-repetition-time, T1-weighted (eg, fast low-angle shot) or short-repetition-time, steady-state (eg, gradient-recalled acquisition in the steady state) sequences have gained some clinical acceptance. T2-weighted imaging still depends largely on two-dimensional (2D), multisection conventional SE or, more prevalently, fast SE methods (2–5).
Multislab 3D fast SE techniques (6,7) offer the potential for high-spatial-resolution imaging of the whole brain but nonetheless possess several important limitations. Radio-frequency (RF) excitation along the section-select direction is nonuniform, and, as a result, 3D reformations generally are contaminated by artifacts at the junctions between slabs. In addition, because of slab profile effects, some of the outer sections in each slab typically are discarded, thus decreasing the efficiency. Power deposition is relatively high and may compromise the coverage attained per unit time, particularly at a field strength of 1.5 T and greater. Each slab undergoes unwanted magnetization-transfer effects from the large number of high-flip-angle off-resonance RF pulses applied to the other slabs during the acquisition (8). The use of additional contrast preparations, such as magnetization-transfer contrast or diffusion, requires that the repetition time be increased or the anatomic coverage compromised.
Recent advances in whole-body MR imager technology, especially higher performance gradient and receiver systems, now present the opportunity for developing single-slab 3D techniques that capitalize on SE acquisitions. In this article, we report our initial efforts to develop and optimize single-slab 3D brain imaging based on fast SE techniques (2–5). The major technical issues that had to be addressed are presented, followed by preliminary imaging results for 3D T1-weighted, T2-weighted, and fluid-attenuated inversion-recovery, or FLAIR, imaging.
|
Materials and Methods |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionREFERENCES |
|---|
Pulse Sequence Design
The major technical issues that arose during the development of the single-slab 3D pulse sequences included receiver-system dynamic range limitations due to the large imaging volume and image ghost artifacts related to the fidelity of the gradient waveforms. Other important sequence design considerations included the configuration of the spoiler (crusher) gradient waveforms and the implementation of a half-Fourier acquisition mode.
Dynamic range limitations.—It is well known that the signal dynamic range required for Fourier-based encoding of large volumes, such as the whole brain, can exceed that provided by the 16-bit analog-to-digital convertors typically used in the receiver systems of MR imagers (10,11,12). With insufficient dynamic range, quantization (digitization) noise can dominate random thermal noise and consequently degrade image quality, primarily by decreasing the image signal-to-noise ratio (SNR). Compared with a thin-section 2D acquisition, a thick 2D section or 3D slab requires that the gain setting of the receiver is lowered to accommodate the larger signal amplitude. As a consequence, the amplitude of the thermal noise from the coil and subject is decreased relative to the peak signal amplitude, and at sufficiently low receiver-gain settings, the amplitude of the noise will be less than the resolution of the digitized signal. For these conditions, the noise intensity is determined by the resolution of the analog-to-digital convertor, not by the inherent thermal noise, and is effectively increased relative to the peak signal amplitude, which consequently reduces the SNR.
Ultimately, whether the thermal noise is digitized adequately depends not only on the resolution of the analog-to-digital convertor but also on the sampling frequency (ie, on the degree of oversampling), the RF coil characteristics, the receiver-gain setting, and the noise performance of the other components in the receiver system. Nonetheless, experimental characterization of the noise as a function of receiver gain can be used to establish whether quantization noise is an issue. To perform such a characterization of the noise for our specific application, the relationship between receiver gain and image noise intensity was determined by acquiring image sets with the standard head RF coil over the relevant range of receiver-gain settings. The noise intensity was measured as the mean of the random noise in the image background within a region of interest that included a minimum of 5,000 pixels and excluded any visible artifacts from motion or data truncation.
Although several methods for overcoming dynamic range limitations have been described (10–13), none has been adopted for widespread use, since the majority of clinical MR imaging applications currently depend on thin-section 2D acquisitions. Of the previously described methods, only a variable gain approach (14,15), wherein the receiver gain is varied during the acquisition to match the signal level, requires no modification of the pulse sequence. To address the dynamic range limitation associated with imaging the whole brain by using a single thick slab, an adaptation of the gain-changing approach was developed as illustrated in Figure 1. This approach takes advantage of the multiple receiver channels available for phased-array RF coils to simultaneously record the signal from the standard head coil at two receiver-gain settings. The low-gain data, which are contaminated by quantization noise, are used for only the central, high-signal portion of spatial-frequency space. In practice, approximately 0.03% of the total number of spatial-frequency-space components in the 3D data set are from the low receiver-gain setting. The reconstruction software is programmed to automatically scale and combine the data from the two receiver channels. This two-receiver-channel method has been integrated into one of the two imagers used in this study (Vision; Siemens Medical Systems, Iselin, NJ) and is currently being developed for the other imager (Signa; GE Medical Systems, Milwaukee, Wis).
|
Gradient-related ghost artifacts.—The high sensitivity of Carr-Purcell-MeiboomGill SE trains, as used in fast SE sequences, to errors in the phase of the transverse magnetization is well documented (16,17). Common sources of these errors include eddy currents and RF-pulse phase instabilities. However, any differences between 2D and 3D sequences in their sensitivity to phase errors has not, to our knowledge, been recognized. Our investigation was prompted by the finding that preliminary versions of the 3D pulse sequences demonstrated substantial ghost artifacts (peak artifact intensity of about 10% of the mean signal intensity within the object of interest), whereas comparable 2D sequences showed negligible ghost artifacts.
For a given sequence configuration, the specific source and form of gradient-induced phase errors is dependent on the details of the gradient-system and eddy-current behaviors associated with a particular imager. For an example of a potential source for gradient-induced phase errors, we chose to study an error in the area of the preread gradient waveform (ie, the gradient waveform applied along the signal readout direction between the excitation and first refocusing RF pulses). Experimental and simulated images from 2D and 3D sequences with identical timing were assessed for the presence and severity of ghost artifacts as a function of an error in the area of the preread gradient waveform that ranged between 0% and 0.2% of the area corresponding to the readout gradient waveform. Simulated images also were used to predict the fraction of image intensity present as ghost artifacts as a function of gradient-area error and refocusing RF-pulse flip angle.
Spoiling configuration.—Spoiler or crusher gradients are used in fast SE pulse sequences between the refocusing RF pulse and data sampling window of each echo to dephase the free induction decay generated by the refocusing RF pulse. These gradient waveforms are often placed along the section-select direction (7). However, in 3D sequences, this configuration may substantially decrease the efficiency of the pulse sequence because data sampling does not begin until both the spoiling and 3D phase-encoding gradients are applied. A more time-efficient approach, illustrated in Figure 2, is simply to extend the duration of the readout gradient waveform a sufficient amount to displace signals originating at the refocusing RF pulses out of the data sampling window. Thus, encoding in the third dimension and spoiling are performed simultaneously, which permits a decreased echo spacing compared with that achieved when spoiling is performed along the section-select direction.
|
The implementation of the half-Fourier acquisition differs in an important way from that used in another recently developed 3D SE-based pulse sequence (19). In our case, a full-Fourier acquisition was obtained along the phase-encoding direction in spatial-frequency space that is modulated by signal decay during the echo train (the in-plane phase-encoding direction), and a half-Fourier acquisition was obtained along the direction of spatial-frequency space that has constant signal-strength weighting (the 3D phase-encoding direction).
This is in contrast to the method used for 3D inversion-recovery half-Fourier single-shot rapid acquisition with relaxation enhancement (19), which places the half-Fourier acquisition along the phase-encoding direction of spatial-frequency space that is modulated by signal decay during the echo train. Thus, with half-Fourier single-shot rapid acquisition with relaxation enhancement, the weighting along the half-Fourier direction of spatial-frequency space has the form of a low-pass filter, which potentially results in blurring for tissues with relatively short T2 relaxation times.
MR Imaging
Imaging was performed by using two 1.5-T commercial whole-body imagers (Vision and Signa). Both MR systems were equipped with echo-planar-capable gradient systems having maximum gradient strengths and minimum rise times of 25 mT/m and 300 µsec (Vision) and 23 mT/m and 300 µsec (Signa). Initial testing and optimization of the sequences were performed by using phantoms that were supplied with the imagers by the manufacturers and that contained water doped with nickel sulfate or nickel chloride. Preliminary images obtained in the heads of healthy volunteers and patients were then acquired by using the optimized sequence configurations after obtaining informed consent. Institutional review board approval was obtained for all human imaging protocols. The pulse sequence parameters used for 3D T1-weighted, T2-weighted, and fluid-attenuated inversion-recovery imaging are summarized in the Table.
|
A consensus evaluation of the images from volunteers and patients was performed by four of the authors (C.R.G.G., J.P.M., R.L.R., R.V.M.). The gross contrast behaviors were assessed in comparison with those established for 2D conventional SE and fast SE imaging, and the images were evaluated for the presence and severity of physiologic (eg, motion) and nonphysiologic artifacts. In this preliminary study, no formal grading of the image characteristics was performed.
|
Results |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionREFERENCES |
|---|
|
|
|
Gradient-related Ghost Artifacts
Images from the 3D pulse sequences demonstrated visible ghost artifacts for preread gradient-area errors of only a few hundredths of a percent, whereas images from 2D pulse sequences in which identical timing parameter values were used showed negligible ghost artifacts over the complete range of gradient-area errors. The only substantial artifact noted in the 2D images was an intensity shading along the readout direction for the larger values of the gradient-area error. Figure 5 shows a comparison of 2D and 3D images for a gradient-area error of 0.1% and a refocusing-RF-pulse flip angle of 180°. The 3D images (Fig 5b, 5d) demonstrate both diffuse and focal degradation compared with the 2D images (Fig 5a, 5c).
|
|
|
|
|
Although 2D images acquired by using a 180° refocusing RF-pulse flip angle were predicted (Fig 6) to yield an artifact fraction only slightly lower than that for 3D images acquired by using a 90° flip angle, the severity of artifacts for the 2D case, on the basis of visual inspection of both simulated and experimental images, was substantially lower. This apparent discrepancy was explored by using simulated images to investigate the nature of the ghost artifacts as a function of the position within the 2D section profile. Evaluation of these images indicated that the low level of ghost artifacts in the 2D case arises for two reasons: (a) The section-selective refocusing RF pulse contains a range of flip angles between 0° and 180°; as the flip angle decreases with distance from the center of the section, so does the relative intensity of the ghost artifacts. (b) The location of the ghost artifacts is a strong function of the flip angle.
The relationship between the location of ghosts and the flip angle was recognized previously by Norris et al (16) in connection with fast 2D rapid acquisition with relaxation enhancement pulse sequences in which low-flip-angle refocusing RF pulses were used. Since the artifact seen in a 2D image is the result of integrating over the section, and hence over the corresponding range of flip angles, there is limited constructive interference among the artifacts arising from different portions of the section. Thus, in contrast to the 3D case, the artifacts are spread more evenly across the image and appear primarily as a low-intensity diffuse smearing that for most practical purposes is irrelevant.
Whole-Brain 3D Imaging Examples
Figures 7 and 8 present sagittal, coronal, and transverse T1- and T2-weighted images reconstructed at 1.0-mm thickness from 9.4-minute 3D acquisitions covering the whole head. For both image sets, the spatial resolution along each dimension was approximately 1 mm. These image sets demonstrate the typical image quality and contrast obtained from the T1- and T2-weighted single-slab 3D acquisitions, and they illustrate the capability to obtain high-quality, high-spatial-resolution image reformations in any plane. Figure 9 presents an additional example of sagittal, coronal, and transverse 3D T1-weighted images, reconstructed at 1.0-mm thickness, in a patient with an optic glioma. This gadolinium-enhanced image set demonstrates high contrast between the tumor and surrounding tissue, which is comparable to that for gadolinium-enhanced 2D T1-weighted conventional SE imaging. Again, with 1-mm spatial resolution along each dimension, high-spatial-resolution images can be obtained in any plane.
|
|
|
|
The only other artifact sometimes noted was aliasing (wraparound) of signal from outside the imaging volume into the images because of a suboptimal choice of parameter values. An example of this is shown in the sagittal image in Figure 9. The combination of a transverse primary orientation and an insufficient slab thickness resulted in some aliasing artifact (eg, in the corpus callosum), even though a spatial-presaturation slab was placed inferior to the imaging volume.
|
Discussion |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionREFERENCES |
|---|
In developing the 3D pulse sequences, we discovered an important difference between 2D and 3D SE-train–based techniques related to gradient-system performance. For the specific type of gradient-waveform error studied, the 2D sequences did not demonstrate substantial shading artifacts until the gradient error was on the order of 0.1%, whereas the 3D sequences demonstrated visible ghost artifacts for a gradient error that was an order of magnitude smaller. Thus, successful implementation of the 3D SE-train sequences requires substantially higher gradient-waveform fidelity than that needed for 2D SE-train sequences. In practice, we found that a combination of precautions was effective in maintaining gradient-related ghost artifacts at a negligible level (ie, peak artifact intensity less than approximately 1% of the mean signal intensity within the object of interest) in the 3D sequences. These included (a) ensuring that eddy-current compensation was within manufacturer specifications, (b) using consistent waveform slew rates and spacing when gradient waveforms were placed immediately adjacent to short-duration RF pulses, and (c) using low-flip-angle refocusing RF pulses for the shortest echo spacings.
The 3D T1-weighted images yielded contrast comparable to that of conventional 2D SE images, but the contrast between brain white and gray matter appeared to be less than that typically achieved by using 3D gradient-recalled- echo techniques such as fast low-angle shot or magnetization-prepared rapid gradient echo. However, for contrast agent–enhanced 3D imaging, the SE-based technique may be preferred, considering the reported limitations of gradient-recalled-echo techniques for depicting contrast-enhanced lesions (23,24). Compared with the current standard for contrast-enhanced imaging, conventional 2D SE, the 3D SE-based technique also provides the advantages of thin contiguous sections and fewer flow-related artifacts, which suggests the potential for the 3D technique to replace conventional 2D gadolinium-enhanced SE imaging.
Despite the promising results presented for single-slab 3D SE-based imaging in this article, the issue of which type of 3D SE-based imaging—single-slab or multislab—will prove to be superior for imaging the whole brain in arbitrary orientations with isotropic resolution is still open to debate. With multislab 3D imaging, increased echo spacing is possible while still providing sufficient coverage to image the whole brain with an acquisition time comparable to that used with single-slab 3D techniques. Thus, the multislab approach potentially can achieve a higher sampling efficiency (the fraction of the echo spacing for which the analog-to-digital convertor is on) and therefore an increased SNR per unit time.
However, multislab implementations typically discard a number of the outer sections from each slab to decrease slab boundary artifacts, which in turn decreases sampling efficiency. In addition, the use of echo trains that combine gradient echoes and SEs with 3D SE-based imaging (25,26) decreases the theoretic sampling efficiency advantage of multislab techniques. Although substantial pulse sequence improvements have been made with regard to reducing slab-boundary artifacts (27,28), these artifacts are yet to be completely eliminated. Recently, a potential alternative, the shifted interleaved multivolume acquisition, has been introduced (29). This method eliminates slab-boundary artifacts in exchange for ghosting in an orthogonal direction. Overall, considering the results presented to date for the single-slab and multislab approaches, further development and direct comparison of both techniques seems warranted.
Although our studies primarily concentrated on obtaining image sets with approximately isotropic 1-mm spatial resolution, different configurations of the single-slab 3D techniques may prove useful for other applications, particularly when high-spatial-resolution image reformations in arbitrary planes are not needed. One example is the 3D fluid-attenuated inversion-recovery images in Figure 10 in which 1.0-mm sections were traded for an extended repetition time. With comparable or increased SNR, the current T1- and T2-weighted sequences could acquire 2.0-mm-thick sections covering the whole head in less than 5 minutes by using 1-mm in-plane resolution or acquire 3.5-mm-thick sections in less than 10 minutes by using 0.5-mm in-plane resolution.
In conclusion, MR pulse sequences for single-slab 3D brain imaging, based on fast SE techniques, have been developed. Optimization of the pulse sequence structure to overcome SNR losses induced by quantization noise, to suppress ghost artifacts secondary to gradient-system performance, and to maximize sampling efficiency permitted acquisition of 3D image sets of the whole brain in less than 10 minutes by using T1- or T2-weighted contrast and a volume resolution of 1 mm3 or fluid-attenuated inversion-recovery contrast and a volume resolution of 3 mm3. The development of these techniques revealed important pulse sequence design issues unique to the single-slab 3D approach. In our preliminary experience, these pulse sequences routinely provided high-quality image sets, free of any substantial image artifacts. These techniques have the potential to improve the state of the art both for routine MR imaging and for specialized applications such as surgical planning or lesion quantification for therapy monitoring. The ability to collect such high-spatial-resolution 3D data sets in a clinically reasonable time will affect the diagnosis, treatment, and research for disseminated diseases of the brain, such as multiple sclerosis, Alzheimer disease, and metastatic neoplasms.
|
FOOTNOTES |
|---|
Author contributions: Guarantors of integrity of entire study, J.P.M., F.A.J.; study concepts, J.P.M., R.V.M., C.R.G.G., F.A.J., J.R.B.; study design, J.P.M.; definition of intellectual content, J.P.M., R.V.M., C.R.G.G., J.R.B.; literature research, J.P.M., R.V.M.; clinical studies, R.L.R., R.V.M., S.B., C.R.G.G.; experimental studies, J.P.M., S.B., R.V.M., C.R.G.G.; data acquisition, J.P.M., S.B., R.V.M., C.R.G.G.; data analysis, J.P.M., S.B.; manuscript preparation, J.P.M.; manuscript editing, J.P.M., R.V.M., C.R.G.G., J.R.B.; manuscript review, R.L.R., F.A.J.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  S. Roosendaal, B Moraal, P. Pouwels, H Vrenken, J. Castelijns, F Barkhof, and J. Geurts Accumulation of cortical lesions in MS: relation with cognitive impairment Multiple Sclerosis, June 1, 2009; 15(6): 708 - 714. [Abstract] [PDF]
|
|
|
|
 |
|
 Y. Kato, S. Higano, H. Tamura, S. Mugikura, A. Umetsu, T. Murata, and S. Takahashi Usefulness of Contrast-Enhanced T1-Weighted Sampling Perfection with Application-Optimized Contrasts by Using Different Flip Angle Evolutions in Detection of Small Brain Metastasis at 3T MR Imaging: Comparison with Magnetization-Prepared Rapid Acquisition of Gradient Echo Imaging AJNR Am. J. Neuroradiol., May 1, 2009; 30(5): 923 - 929. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B. Moraal, D. S. Meier, P. A. Poppe, J. J. G. Geurts, H. Vrenken, W. M. A. Jonker, D. L. Knol, R. A. van Schijndel, P. J. W. Pouwels, C. Pohl, et al. Subtraction MR Images in a Multiple Sclerosis Multicenter Clinical Trial Setting Radiology, February 1, 2009; 250(2): 506 - 514. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. J. Stevens, R. F. Busse, E. Han, A. C. S. Brau, P. J. Beatty, C. F. Beaulieu, and G. E. Gold Ankle: Isotropic MR Imaging with 3D-FSE-Cube--Initial Experience in Healthy Volunteers Radiology, December 1, 2008; 249(3): 1026 - 1033. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. K. Kuhl, F. Traber, and H. H. Schild Whole-Body High-Field-Strength (3.0-T) MR Imaging in Clinical Practice * Part I. Technical Considerations and Clinical Applications Radiology, March 1, 2008; 246(3): 675 - 696. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
Y. Duan, P.G. Hildenbrand, M.P. Sampat, D.F. Tate, I. Csapo, B. Moraal, R. Bakshi, F. Barkhof, D.S. Meier, and C.R.G. Guttmann Segmentation of Subtraction Images for the Measurement of Lesion Change in Multiple Sclerosis AJNR Am. J. Neuroradiol., February 1, 2008; 29(2): 340 - 346. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. E. Gold, R. F. Busse, C. Beehler, E. Han, A. C. S. Brau, P. J. Beatty, and C. F. Beaulieu Isotropic MRI of the Knee with 3D Fast Spin-Echo Extended Echo-Train Acquisition (XETA): Initial Experience Am. J. Roentgenol., May 1, 2007; 188(5): 1287 - 1293. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R J Mills, C A Young, and E T S Smith 3D MRI in multiple sclerosis: a study of three sequences at 3 T Br. J. Radiol., May 1, 2007; 80(953): 307 - 320. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. J. W. Pouwels, J. P. A. Kuijer, J. P. Mugler III, C. R. G. Guttmann, and F. Barkhof Human Gray Matter: Feasibility of Single-Slab 3D Double Inversion-Recovery High-Spatial-Resolution MR Imaging Radiology, December 1, 2006; 241(3): 873 - 879. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. J. G. Geurts, P. J. W. Pouwels, B. M. J. Uitdehaag, C. H. Polman, F. Barkhof, and J. A. Castelijns Intracortical Lesions in Multiple Sclerosis: Improved Detection with 3D Double Inversion-Recovery MR Imaging Radiology, July 1, 2005; 236(1): 254 - 260. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
I L. Tan, R. A. van Schijndel, P. J. W. Pouwels, H. J. Ader, and F. Barkhof Serial Isotropic Three-Dimensional Fast FLAIR Imaging: Using Image Registration and Subtraction to Reveal Active Multiple Sclerosis Lesions Am. J. Roentgenol., September 1, 2002; 179(3): 777 - 782. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. F. Kallmes, F. K. Hui, and J. P. Mugler III Suppression of Cerebrospinal Fluid and Blood Flow Artifacts in FLAIR MR Imaging with a Single-Slab Three-dimensional Pulse Sequence: Initial Experience Radiology, October 1, 2001; 221(1): 251 - 255. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
= Vision, 
= Signa. Owing to dynamic range limitations, the noise intensity increased above the lower limit determined by thermal noise as the receiver gain decreased. Thus, for single-slab 3D acquisitions, which typically require relative gain settings of less than 20 dB, the image SNR may be decreased substantially compared with the ideal case of an unlimited dynamic range. Error bars (± 1 SD) from 10 separate experiments, each measuring a complete noise-versus-gain data set, fall within the plotting symbols. To permit direct comparison of the general noise-versus-gain characteristics for the two MR systems, the noise intensity values from each imager were scaled into the same arbitrary range, and an offset was added to each set of absolute receiver-gain values so that lowest receiver gain for each data set was 0 dB.
