| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Clinical Magnetic Resonance Group, Royal Marsden NHS Foundation Trust, Royal
Marsden Hospital, Downs Rd., Sutton, Surrey SM2 5PT, United Kingdom.
2 Department of Statistics, Royal Marsden NHS Foundation Trust, Sutton, Surrey,
United Kingdom.
3 Institute of Cancer Research, Royal Marsden NHS Foundation Trust, Sutton,
Surrey, United Kingdom.
Received December 21, 2005;
accepted after revision June 28, 2006.
Supported by Cancer Research UK (grant number CUK C1060/A808). S. A.
Reinsberg was supported by Department of Health UK, New and Emerging
Applications of Technology, grant B132.
Abstract
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
SUBJECTS AND METHODS. Forty-two men (mean age ± SD, 69.3
± 4.7 years) with prostate cancer were studied using endorectal
T2-weighted imaging, 2D chemical shift imaging (CSI), and isotropic apparent
diffusion coefficient (ADC) maps. Regions of interest (ROIs) were drawn around
the entire gland, central gland, and peripheral zone tumor, diagnostically
defined as low signal intensity on T2-weighted images within a sextant that
was biopsy-positive for tumor. Lack of susceptibility artifact on a
gradient-echo B0 map through the slice selected for CSI and no high
signal intensity on external array T1-weighted images confirmed the absence of
significant hemorrhage after biopsy. CSI voxels were classified as
nonmalignant or as tumor (ROI included 
30% or 3 70% tumor).
Choline-citrate (Cho/Cit) ratios and average ADCs were calculated for every
voxel. A plot of Cho/Cit ratios versus ADCs yielded a line of best separation
of tumor voxels from nonmalignant voxels. Receiver operating characteristic
(ROC) curves were plotted for Cho/Cit ratios alone, ADCs alone, and a
combination of the two.
RESULTS. The Cho/Cit ratios were significantly higher (p < 0.001) and the ADCs were significantly lower (p < 0.006) in tumor-containing voxels than in non-tumor-containing voxels. When voxels containing 30% or more tumor were considered positive, the area under the ROC curves using combined MR spectroscopy and ADC (0.81) was similar to that of Cho/Cit alone (0.79) and better than ADC alone (0.66). When voxels containing 70% or more tumor were considered positive and cutoffs to achieve a 90%-or-greater sensitivity chosen, a combination of Cho/Cit and ADC achieved a significant improvement in specificity compared with Cho/Cit alone (p < 0.0001) or ADC alone (p < 0.0001).
CONCLUSION. When voxels containing 70% tumor are considered
positive, the combined use of MR spectroscopy and diffusion-weighted MRI
increases the specificity for prostate cancer detection while retaining the
sensitivity compared with MR spectroscopy alone or diffusion-weighted MRI
alone.
Keywords: diffusion-weighted MRI • genitourinary tract imaging • MRI • MR spectroscopy • MR technique • oncologic imaging • prostate cancer
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
MR spectroscopy offers some improvement in sensitivity and specificity for prostate cancer detection [4-6]. Specifically, a significant reduction of citrate (Cit) and an elevation of choline (Cho) levels have been documented in prostatic adenocarcinomas as compared with normal prostatic peripheral zone tissue and benign prostatic hypertrophy. Recently, the use of 3D chemical shift imaging (CSI) has been shown to improve the sensitivity and specificity for tumor detection to 95% and 81%, respectively [7]. Another alternative to T2-weighted MRI is to develop image contrast through "apparent diffusivity," tissue water incoherent displacement over distances of approximately 1-20 µm. Diffusion-weighted MRI is showing potential for improving prostate cancer detection [8-10]. The purposes of this study therefore were to determine the metabolic ratios on 2D chemical shift 1H MR spectroscopy and the apparent diffusion coefficient (ADC) in patients with histologically proven prostate cancer and to establish the sensitivity and specificity of prostate cancer detection using a combined MR spectroscopy and diffusion-weighted MRI approach.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
A minimum of six cores had been obtained from each patient. The number of positive cores varied from one to four, and Gleason grade varied from 6 to 8. Patients ranged in age from 60 to 78 years (mean ± SD, 69.3 ± 4.7 years) with PSA values ranging from 0.45 to 45 ng/mL (median, 10.2 ng/mL; lower and upper quartiles, 6.0 and 16.2 ng/mL). Over a 13-month period, patients underwent diffusion-weighted MRI and 1H MR spectroscopy of the prostate during the same examination as the routine staging MRI examination performed at a median of 21 days (lower and upper quartiles, 12 and 49 days) after the most recent biopsy.
MRI and 1H MR Spectroscopy
MR studies were performed on a 1.5-T unit (Intera, Philips Medical Systems)
using an endorectal coil with a balloon inflated with 55 mL of air. Hyoscine
butyl bromide (20 mg) was administered intramuscularly immediately before
centering the patient in the scanner to reduce peristalsis: The use of
hyoscine butyl bromide is routine at our institution for abdominopelvic MRI,
and hyoscine butyl bromide is preferred to glucagon because of its more
effective antiperistalsis capability. None of our patients had a history of
urine retention. Although hyoscine butyl bromide is contraindicated in
patients with large prostates and urine retention, given intramuscularly at
this dose, we have seen no cases of urine retention in the more than 500
prostate examinations we have performed.
Conventional T2-weighted turbo spin-echo (TSE) images were obtained in three orthogonal planes (TSE; TR/effective TE, 2,000/90; echo-train length, 16; 2 signal averages) with a 256 x 512 matrix, 3-mm slice thickness, no gap, and a 14-cm field of view; the total imaging time was 12 minutes. Echo-planar diffusion-weighted images (TR/TE, 2,500/69) with b values of 0, 300, 500, and 800 s/mm2 were obtained transverse to the prostate and parallel to the corresponding set of T2-weighted images. The phase-encoding gradient was from left to right to minimize motion artifacts in the prostate. Twelve 4-mm-thick slices (no gap; 20-cm field of view; matrix, 128) provided coverage of the prostate with an image acquisition time of 1 minute 24 seconds. ADC maps were generated using the system software.
In addition, CSI was performed over a single slice transverse to the prostate, using a thickness of 15 mm and a 16 x 16 grid (voxel size, 8.75 x 8.75 x 15 mm3). The slice was selected by identifying a focal low-signal-intensity abnormality on the axial T2-weighted scans within a sextant that was biopsy-positive for tumor and centering there. Signal collection was restricted to voxels over the prostate using the point-resolved spectroscopy (PRESS) localization technique (TR/TE, 1,500/120). An automated shimming method over the PRESS volume used scanner software to adjust the x, y, and z axis gradients in small increments and test for the best line shape. Lipid suppression, by frequency-selective inversion before PRESS excitation, and water suppression, using band-selective inversion with gradient dephasing (BASING [11]), were included. Regional saturation was not required due to acceptably low levels of lipid contamination in both the central and peripheral voxels. Data acquisition took 12 minutes. The spectroscopy data were voxel-shifted to align the CSI data set with the PRESS box and were exported for analysis with LCModel fitting package (Stephen Provencher, Inc.) [12] using a basis set containing metabolite spectra from choline, creatine, and citrate. B0 maps (dual gradient-echo; TR/first-echo TE, second-echo TE, 50/7.6, 17.6) were obtained before MR spectroscopy in all cases to ensure that there was no significant susceptibility artifact from hemorrhage within the prostate.
After endorectal MRI and MR spectroscopy, an external pelvic phased-array coil was used to acquire axial T1-weighted (TSE; TR/TE, 650/7) and T2-weighted (TSE; TR/effective TE, 5,396/80) images through the pelvis as part of the routine clinical staging scan. The B0 maps and T1-weighted images confirmed the absence of any significant postbiopsy hemorrhage in this patient cohort.
Data Analysis
The five consecutive T2-weighted slices that corresponded to the thicker
CSI slice were identified. On each of these, ROIs were drawn by a radiologist,
who had 8 years' experience in interpreting prostate MRI at the time of the
study, for the entire prostate; central gland; and tumor, defined as
low-signal-intensity areas within the peripheral zone in a sextant that was
biopsy-positive for tumor. Isointense tumors were thus not included in this
analysis. The median tumor ROI size was 1.45 cm2, with lower and
upper quartiles of 0.69 and 2.1 cm2. The absence of significant
postbiopsy hemorrhage was confirmed on the B0 maps and T1-weighted
external array images.
The first criterion chosen for inclusion of a voxel in the analysis was that the overall signal-to-noise ratio (SNR) calculated by the LCModel fitting package [12] in each spectrum should be greater than 2. The SNR in the LCModel is defined as the maximum signal within the frequency range of interest in the baseline-corrected spectrum divided by twice the root-mean-square residuals. A possible alternative criterion that both the choline and citrate signals must satisfy a specific Cramer-Rao minimum variance bound would result in voxels containing significant tumor volume being rejected due to an absence of citrate. The criterion chosen led to the inclusion of spectra from a total of 511 voxels in 42 patients. A second criterion was that the voxel included should be formed of at least 70% prostate gland as determined by the ROIs in the five corresponding T2-weighted slices. This process eliminated an additional 113 voxels, leaving a total of 398 voxels for inclusion in the analysis.
Tumor voxels were defined using two different thresholds: 30% or
70% tumor based on proportion of tumor ROI identified on the T2-weighted image
within the voxel. Using the biopsy-supported ROIs outlined on the T2-weighted
scans, the 398 CSI voxels were classified into nontumor peripheral zone (0%
tumor and 70% peripheral zone, n = 24), nontumor central gland
(0% tumor and 70% central gland, n = 180), nontumor mixed voxels
(0% tumor, < 70% central gland, < 70% peripheral zone, n = 68),
and tumor (tumor ROI occupying 30% [n = 47] or 70%
[n = 15] of the voxel). The small number of nontumor peripheral zone
voxels is due to central gland enlargement in this age group, resulting in
predominantly central gland or mixed voxels. Thus, 319 and 287 of the 398
voxels were included in the 30% tumor and 70% tumor analyses,
respectively; the remaining voxels contained smaller tumor volumes and were
not considered. ADC maps were coregistered with high-resolution T2-weighted
images and CSI spectroscopy. Metabolite levels and their ratios and average
ADCs were calculated for every voxel classified on MR spectroscopy (Figs.
1A,
1B,
1C,
1D,
1E, and
1F). The LCModel fitting
package [12] calculated the
Cramer-Rao minimum variance bounds (% SD) for each peak; a peak was regarded
as significant if this value was less than 20%.
|
|
|
|
|
|
Statistical Analysis
Statistical analysis of the data was performed using SPSS software (version
11.0, SPSS) for Windows (Microsoft) and in-house software developed using
Interactive Data Language (IDL, ITTVIS Ltd.). Because clear separation between
choline and creatine was obtained in our data, the ratios of Cho/Cit (rather
than needing to use [choline + creatine] / citrate) and ADC from each voxel
type (nontumor, 30% tumor, and 70% tumor) were computed and
compared. Normality plots and the Shapiro and Francia test for normality
showed the data from the Cho/Cit ratios were nonparametric, whereas the data
from the ADC maps were parametric. Therefore, the Cho/Cit data were normalized
by taking the log (to base 10) of values, and both sets of data were
investigated using parametric statistical tests. One-way analysis of variance
tests were used to detect differences between nontumor and either 30%
tumor or 70% tumor voxels. The Student's t test for independent
samples was used to compare peripheral zone or central gland voxels with tumor
voxels. Significance tests were two-sided, and a p value of less than
0.05 was chosen as the criterion for statistical significance.
|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
The Cho/Cit ratio was higher in tumor-containing voxels than in
non-tumor-containing voxels (Table
1). Univariate analysis of variance for each tumor threshold
showed that the Cho/Cit ratios were significantly higher in both classes of
tumor-containing voxels than in non-tumor-containing voxels (p <
0.001). When 30% tumor within a voxel was considered positive, an ROC
curve of the relation between the log10 (Cho/Cit) and the diagnosis
of tumor showed the AUC to be 0.79 (95% CI, 0.71-0.88;
Fig. 3A). For this 30%
threshold, the optimum discrimination suggested by the ROC curve was at a
cutoff Cho/Cit ratio of > 0.07 (log10[Cho/Cit] = -1.15); voxels
containing 30% tumor could be predicted with a sensitivity of 80.9% and
specificity of 64.3% (Table 2).
When 70% tumor within a voxel was considered positive, the optimum
discrimination was at a cutoff value of Cho/Cit ratio of > 0.081
(log10[Cho/Cit] = -1.089), resulting in a sensitivity of 93.3% and
specificity of 73.2% (AUC, 0.92; 95% CI, 0.85-0.99;
Fig. 3B). This improvement was
statistically significant (p < 0.01).
|
|
|
|
Mean ADC values were obtained from voxels defined for 2D MR spectroscopy. Because of the large voxel size, the average values obtained did not differentiate tumor voxels with the sensitivity that is possible when using ADC maps at their native resolution of 1.5 mm3 (Figs. 1A, 1B, 1C, 1D, 1E, and 1F). ADC values were lower in tumor-containing voxels than in non-tumor-containing voxels (Table 1). Voxels in the central gland tended to show smaller ADCs than peripheral zone voxels and to partially overlap with voxels comprising 30-70% tumor (Table 1).
Univariate analysis of variance showed that the ADCs were significantly
lower in both classes of tumor-containing voxels than in the
non-tumor-containing voxels (p < 0.006). There were also
significant differences between normal peripheral zone and central gland
voxels (p < 0.001) and between voxels containing 30% tumor
and those containing 70% tumor (p < 0.01). When 3
30% tumor within a voxel was considered positive, an ROC curve of the relation
between the ADC and the diagnosis of tumor showed the AUC to be 0.66 (95% CI,
0.58-0.75; Fig. 3A). No cutoff
value of ADC could be chosen to achieve at least 70% sensitivity and
specificity. When voxels containing 70% tumor were considered, the AUC
was 0.85 (95% CI, 0.76-0.95) and using a cutoff of 1.26 x
10-3 mm2/s, tumor could be predicted with a sensitivity
of 93.3% and specificity of 57.4% (Table
2). Sensitivity and specificity for 30% tumor voxels using
this ADC cutoff were 67.3% and 59.9%, respectively.
A combined analysis of Cho/Cit ratios and ADCs was also done. When
3 30% tumor within a voxel was considered positive, an ROC curve of
the relation between a combination of log10(Cho/Cit) and ADC and
the diagnosis of tumor showed the AUC to be 0.81 (95% CI, 0.73-0.88), and
there was no improvement compared with the Cho/Cit curve alone
(Fig. 3A). When 70% tumor
within a voxel was considered positive, an ROC curve of the relation between a
combination of log10 (Cho/Cit) and ADC and the diagnosis of tumor
showed the AUC to be 0.98 (95% CI, 0.95-0.995)
(Fig. 3B and
Table 2), which achieved a
better separation of tumor-containing and non-tumor-containing voxels than ADC
alone or log10(Cho/Cit) alone (p < 0.05). The optimal
separation for the 70% tumor-containing voxels is described by the
relationship log10(Cho/Cit) = 1.946 x ADC - 3.088
(Fig. 2). Using this dividing
line gives a sensitivity of 93.3% and a specificity of 90.9% for tumor
diagnosis.
For comparison, a similar optimized separating line for the 30% tumor
voxels given by the relationship log10(Cho/Cit) = 0.910 x ADC
- 2.102 resulted in a sensitivity of 70.2% and a specificity of 83.5%. Because
achieving > 90% sensitivity was used as the criterion for comparison, it
was not possible to test for an improvement in sensitivity. However, for
70% tumor-containing voxels, specificity of the combination was significantly
improved compared with either Cho/Cit ratio alone (p < 0.0001) or
ADC alone (p < 0.0001). With a sensitivity of 100% for the
combination of Cho/Cit and ADC, a specificity of 90% could be achieved
(Table 2).
To assess the quality of the ROC curves used for predicting the sensitivity and specificity tumor presence, the AUC was plotted against tumor fraction in the voxel (Fig. 4). The AUC values increased with increasing tumor fraction in the voxel, consistent with a reduced contamination of the data from nonmalignant tissue within the voxel.
|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
70% tumor are considered
positive for tumor, the combination of MR spectroscopy and diffusion-weighted
MRI increases the specificity of prostate cancer detection while retaining
sensitivity compared with MR spectroscopy alone or diffusion-weighted MRI
alone. The altered cellular metabolites from tumor-containing voxels and
changes in ADC reflecting restricted extracellular water diffusion in
hypercellular tumor regions were considered together to produce significant
increases in specificity compared with either technique alone. Although
endorectal T2-weighted MRI is well established as a local staging technique
for identifying extracapsular tumor extension
[13,
14], newer treatment
strategies (e.g., imaging-guided brachytherapy
[15,
16], photodynamic therapy
[17], and cryoablation
[18]) demand more precise
information about the presence of tumor and the location of tumor in
organ-confined disease in order for treatment areas to be tailored to regions
of the prostate gland where cancer is present. Moreover, patient selection for
and maintenance in an active surveillance program are assisted by accurate
information about tumor localization, size, and growth and may obviate repeat
biopsies.
In patients with an elevated PSA level but repeatedly negative findings at prostatic biopsy, identifying likely sites of tumor is vital for guiding future biopsies. T2-weighted endorectal MRI alone has proved useful in reliably depicting large (> 1 cm) tumors [2], but remains prohibitive for diagnosing small tumors. We have shown the potential of a combined MR spectroscopy and diffusion-weighted imaging approach for prostate cancer detection. However, to define the best cutoff values for metabolite ratios and ADCs, further work is needed to determine their reproducibility. Also, the voxels used in our study were relatively large; future use of 3D MR spectroscopy techniques with smaller voxels would permit the detection of smaller tumors. It would also make the addition of diffusion-weighted imaging more valuable because diffusion-weighted imaging would have fewer partial volume effects from being used at a large spectroscopic voxel resolution.
The use of MR spectroscopy as a diagnostic tool for improving the accuracy of prostate cancer detection has been extensively investigated [4, 7, 19]. Initial studies using multivoxel MR spectroscopy with T2-weighted MRI showed that the combination improved accuracy of cancer localization to 81% compared with 71.4% for MRI alone [20]. Studies for evaluating sensitivity and specificity have largely been limited by the availability of detailed histologic comparison at radical prostatectomy. One study of 47 patients comparing MRI and MR spectroscopy with sextant biopsy against step section histology showed that MRI and MR spectroscopy were more sensitive but less specific than sextant biopsy at prostate cancer detection [21] and that the combination was better than MRI alone. With the use of various combinations of MRI and 3D MR spectroscopic imaging, a point on the ROC curve can be chosen that provides either high sensitivity or high specificity depending on clinical requirements. A positive result with combined MRI and 3D MR spectroscopic imaging (> 3 SDs) indicates the presence of tumor with high probability (positive predictive value, 89-92%), whereas a negative result (> 2 SDs) excludes the presence of cancer with high probability (negative predictive value, 74-82%). The main limitation of our study for clinical application is the acquisition of only a single-slice 2D CSI. However, this does not affect the validity of our finding and is not a fundamental limitation; both multislice 2D CSI and full 3D CSI can also be implemented.
In our study, water suppression using BASING [11], together with the further reduction of residual water within the LCModel, was sufficient to produce spectra with negligible interference from water. Some of the early prostate MR spectroscopy measurements used both CHESS (chemical shift selective) for water suppression and inversion pulses for fat suppression, a combination that inevitably led to some compromises and imperfect suppression of water and lipid. Dual-band BASING and spectral-spatial pulses have been recommended to solve this problem [22]. However, we found that the combination of a simple frequency-selective inversion pulse followed by a relaxation delay before the PRESS localization reduced signals from lipid to a satisfactory level. Also, although some authors have found it helpful to use saturation bands to reduce lipid contamination from outside the slice [22], visual inspection of our data showed no sign of large lipid peaks. In addition, in two patients scanned with and without saturation bands, a comparison of 40 voxels (with sufficient SNR to give < 20% uncertainty in peak areas) showed no significant difference (p > 0.1) in mean metabolite values.
The sensitivity of MR spectroscopy, however, may be affected by cancer grade: A study by Zakian et al. [23] suggests that small low-grade tumors may be missed using 3D MR spectroscopic imaging because the severity of metabolite alteration correlates with tumor aggressiveness. There was a trend toward increasing the ([Cho + creatine] / Cit) value with increasing Gleason score in lesions identified correctly with MR spectroscopic imaging. In our study, all but one of the tumors had a Gleason grade > 6.
More recently, diffusion-weighted MRI has also been evaluated for its ability to show the prostate, primarily in animal models. An increase in the ADC has been noted in a subcutaneously implanted rat prostate tumor after chemotherapy [24]. Using nuclear MR microscopy, Artemov et al. [25] showed different diffusion rates for two prostate cancer cell lines. A study at 4.7 T showed improved prostate tumor detection in transgenic mice using diffusion-weighted imaging compared with T2 mapping [26]. Preliminary work on diffusionweighted imaging of the prostate in vivo in humans has shown differences between tissue types warranting clinical investigation [8, 9, 27, 28], and a recent study of 60 patients showed that the addition of diffusion-weighted imaging to conventional T2-weighted imaging significantly improved tumor detection [8].
In the present study, we were assessing diffusion-weighted imaging at a much lower resolution with a voxel size commonly used for MR spectroscopy. Despite this, the addition of the diffusion-weighted imaging data to the MR spectroscopy data significantly improved specificity to 90%. Because it is easy to implement, available on most standard MR scanners, and incurs little time penalty (1.5 minutes), it is worth incorporating diffusion-weighted imaging into the standard prostate MRI and MR spectroscopy protocol. In this study, tumor foci of 1 cm3 within a voxel could be identified with a high degree of certainty.
The use of an air-filled balloon endorectal coil potentially has several technical limitations. The local field inhomogeneities cause difficulties in shimming and result in line broadening. Further, our diffusion-weighted images were acquired with an echo-planar readout, which leads to some distortion in the neighborhood of tissue boundaries where there is a discontinuity in magnetic susceptibility. There is also sometimes a shift in the phase-encode direction owing to a slight offset of the water resonance frequency. We have previously quantified both distortions compared with a standard spin-echo readout, which yielded a bulk shift of 0.11-4.28 mm (median, 1.10 mm) and a residual disagreement between echo-planar and T2-weighted prostate outlines of 1.2 ± 0.5 mm (mean ± SD) [27]. These are small compared with the spectroscopic voxel size.
Although it would have been preferable to perform MR spectroscopy 6-8 weeks after biopsy, the examination was done as an addition to the staging MRI scan. This was required within a shorter time frame to implement clinical management. A further limitation of this study is that malignant lesions isointense to normal peripheral zone on T2-weighted images were not included in the analysis. It was not possible to include those regions without histologic proof of their involvement with tumor. Similarly, it was not possible to include tumor within the central gland because differentiation of benign from malignant low-signal-intensity regions was not possible. The error inherent in sampling on endorectal sonographically guided biopsy would make it difficult to place ROIs over a described location with any degree of certainty. There is also a false-positive rate associated with other benign abnormalities on T2-weighted MRI—conditions including prostatitis, fibrosis, intraglandular dysplasia, and glandular atrophy—that also are low in signal intensity and may mimic cancer [29]. To further establish the sensitivity and specificity of the technique, therefore, correlation with prostatectomy specimens is warranted.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  H. K. Lim, J. K. Kim, K. A. Kim, and K.-S. Cho Prostate Cancer: Apparent Diffusion Coefficient Map with T2-weighted Images for Detection--A Multireader Study Radiology, November 18, 2008; (2008) 2501080207. [Abstract] [Full Text]
|
|
|
|
|
|
E. M. Charles-Edwards, C. Messiou, V. A. Morgan, S. S. De Silva, N. A. McWhinney, M. Katesmark, A. D. Attygalle, and N. M. deSouza Diffusion-weighted Imaging in Cervical Cancer with an Endovaginal Technique: Potential Value for Improving Tumor Detection in Stage Ia and Ib1 Disease Radiology, November 1, 2008; 249(2): 541 - 550. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. Tamada, T. Sone, Y. Jo, A. Yamamoto, T. Yamashita, N. Egashira, S. Imai, and M. Fukunaga Prostate Cancer: Relationships between Postbiopsy Hemorrhage and Tumor Detectability at MR Diagnosis Radiology, August 1, 2008; 248(2): 531 - 539. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 N VAN AS, E CHARLES-EDWARDS, A JACKSON, S JHAVAR, S REINSBERG, N DESOUZA, D DEARNALEY, M BAILEY, A THOMPSON, T CHRISTMAS, et al. Correlation of diffusion-weighted MRI with whole mount radical prostatectomy specimens Br. J. Radiol., June 1, 2008; 81(966): 456 - 462. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. K. Kuhl, F. Traber, J. Gieseke, W. Drahanowsky, N. Morakkabati-Spitz, W. Willinek, M. von Falkenhausen, C. Manka, and H. H. Schild Whole-Body High-Field-Strength (3.0-T) MR Imaging in Clinical Practice * Part II. Technical Considerations and Clinical Applications Radiology, April 1, 2008; 247(1): 16 - 35. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
Y. Mazaheri, A. Shukla-Dave, H. Hricak, S. W. Fine, J. Zhang, G. Inurrigarro, C. S. Moskowitz, N. M. Ishill, V. E. Reuter, K. Touijer, et al. Prostate Cancer: Identification with Combined Diffusion-weighted MR Imaging and 3D 1H MR Spectroscopic Imaging--Correlation with Pathologic Findings Radiology, February 1, 2008; 246(2): 480 - 488. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 V. Kundra, P. M. Silverman, S. F. Matin, and H. Choi Imaging in Oncology from The University of Texas M. D. Anderson Cancer Center: Diagnosis, Staging, and Surveillance of Prostate Cancer Am. J. Roentgenol., October 1, 2007; 189(4): 830 - 844. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. J. Stanley The Challenge of Managing Prostate Cancer Am. J. Roentgenol., January 1, 2007; 188(1): 1 - 1. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
). Line optimally separating tumor
from nontumor voxels is shown.
