Radiology
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online before print September 11, 2007, 10.1148/radiol.2451061444
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow All Versions of this Article:
2451061444v1
245/2/507    most recent
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Scheenen, T. W. J.
Right arrow Articles by Heerschap, A.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Scheenen, T. W. J.
Right arrow Articles by Heerschap, A.
(Radiology 2007;245:507-516.)
© RSNA, 2007

Genitourinary Imaging

Three-dimensional Proton MR Spectroscopy of Human Prostate at 3 T without Endorectal Coil: Feasibility1

Tom W. J. Scheenen, PhD, Stijn W. T. P. J. Heijmink, MD, Stefan A. Roell, PhD, Christina A. Hulsbergen–Van de Kaa, MD, PhD, Ben C. Knipscheer, MD, J. Alfred Witjes, MD, PhD, Jelle O. Barentsz, MD, PhD, and Arend Heerschap, PhD

1 From the Departments of Radiology (T.W.J.S., S.W.T.P.J.H., J.O.B., A.H.), Pathology (C.A.H.), and Urology (B.C.K., J.A.W.), Radboud University Nijmegen Medical Centre, Geert Grooteplein 10, 6525 GA Nijmegen, the Netherlands; and Siemens Medical Solutions, Erlangen, Germany (S.A.R.). From the 2005 RSNA Annual Meeting. Received August 21, 2006; revision requested October 24; revision received December 13; accepted January 16, 2007; final version accepted March 1. Supported by the Dutch Cancer Society (KUN 2003-2925). Address correspondence to T.W.J.S. (e-mail: T.Scheenen{at}rad.umcn.nl).


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 ADVANCES IN KNOWLEDGE
 IMPLICATION FOR PATIENT CARE
 References
 
Purpose: To evaluate sensitivity and specificity of proton magnetic resonance (MR) spectroscopy of the prostate with external surface coil elements at 3 T for differentiation of cancer from healthy tissue within an acceptable measurement time, by using histopathologic findings as the reference standard.

Materials and Methods: The study was approved by the institutional review board; informed consent was obtained. Forty-five men (age range, 51–70 years) underwent 3-T MR imaging with external radiofrequency surface coils for signal reception. MR spectroscopy was performed with acquisition-weighted three-dimensional water- and lipid-suppressed point-resolved spectroscopy pulse sequence. Voxels were classified into healthy peripheral zone, central gland, and periurethral zone and cancer tissue. Cancer voxels were classified according to cancer size and certainty in matching histopathologic findings with MR images. After visual inspection of automated fitting of classified voxels, the choline plus creatine–to-citrate (Cho + Cr/Cit) ratio was calculated for all tissues. Area under the receiver operating characteristic curves (Az) values were used to assess accuracy of discrimination of cancer from healthy tissues. P < .05 indicated a significant difference.

Results: After exclusion of four patients with no voxels that passed visual inspection of the automated fit, a median of 82% of the classified voxels per patient was used in the analysis. Mean Cho + Cr/Cit ratios for healthy tissues were 0.22 ± 0.12 (standard deviation) for peripheral zone, 0.34 ± 0.14 for central gland, and 0.36 ± 0.20 for periurethral area; all were significantly different from that of cancer (P < .001). Az for discrimination of probable and definite cancer tissue from healthy tissue for the peripheral zone (0.84) was significantly higher than that for the central gland (0.69) (P < .05).

Conclusion: Three-dimensional proton MR spectroscopy of the prostate, with a combination of only external radiofrequency surface coils at 3 T, can be used to discriminate cancer from healthy tissue.

© RSNA, 2007


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 ADVANCES IN KNOWLEDGE
 IMPLICATION FOR PATIENT CARE
 References
 
Proton magnetic resonance (MR) spectroscopy of the human prostate has proved to be a valuable addition to conventional MR imaging in the localization of prostate cancer (1). Signals of citrate, creatine, and choline can be detected throughout the prostate, with increased levels of choline and decreased levels of citrate being indicative of cancer tissue (2,3). A major issue in MR spectroscopy of the prostate is the use of an endorectal coil, either rigid or inflatable. Apart from the discomfort to the patient caused by this coil, its insertion and the necessary evaluation of its position take up valuable MR imaging time. Although the local sensitivity reached with the use of an endorectal coil can never be reached by using only conventional external array coils, it is worthwhile to explore the possibilities of not using the endorectal coil: hydrogen 1 (1H) MR spectroscopic examinations without an endorectal coil would be easier, faster, less expensive, and truly noninvasive. It would make this technique eligible for more widespread use in a clinical environment.

MR spectroscopy of the prostate without an endorectal coil has previously been performed at a magnetic field strength of 1.5 T with a single spine-array coil element (4). The possibility of coherently adding the signals from multiple coil elements into one spectrum for every voxel of a spectroscopic imaging grid (5) has enabled the use of a combination of multiple coils at 1H MR spectroscopy of the prostate. The drawback of using surface coils is their larger size and further distance from the prostate, which leads to a decrease in signal-to-noise ratio. By using 3 T, part of this loss in signal-to-noise ratio is recovered. The optimal timing for 1H MR spectroscopic imaging by using point-resolved spatially localized spectroscopy (PRESS) (6) of the prostate at 3 T has been presented recently, which makes the technique suitable for use in patients (7). Because the chemical shift (in hertz) of different resonances increases linearly with magnetic field strength, one would expect better separation of the creatine and choline methyl-proton resonances than that in spectra obtained at a field strength of 1.5 T. However, resonances of strongly coupled proton spins from polyamines (spermine, spermidine, among others) or ethanolamine and phosphoethanolamine could be present between 3.0 and 3.3 ppm (8), which prevents the spectrum from reaching the spectral baseline between the creatine and choline signals. Thus, the purpose of our study was to evaluate the sensitivity and specificity of 1H MR spectroscopy of the prostate performed with a combination of external surface coil elements at 3 T for differentiation of cancer from healthy tissue throughout the prostate within a clinically acceptable measurement time, by using histopathologic findings as the reference standard.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 ADVANCES IN KNOWLEDGE
 IMPLICATION FOR PATIENT CARE
 References
 
One author (S.A.R.) is employed by Siemens Medical Solutions (Erlangen, Germany). All other authors had control of inclusion of any data and information that might present a conflict of interest.

Patients
The study was approved by our institutional review board, and written informed consent was obtained from all patients. From September 2004 to December 2005, 45 consecutive men with biopsy-proved prostate cancer who underwent radical prostatectomy within 6 weeks after MR imaging and 1H MR spectroscopic imaging at 3 T were eligible for inclusion in our study. A patient exclusion criterion was previous hormonal therapy. Patient age ranged from 51 to 70 years (mean age, 60 years). Mean prostate-specific antigen level was 9.6 ng/mL (range, 3.6–79.2 ng/mL), and the mean Gleason score, based on findings in prostatectomy specimens, was 6.6 (range, 5–9). The median time between transrectal ultrasonographically guided eight-core sextant biopsy and MR examination was 107 days (range, 21–246 days), and that between MR examination and surgery was 7 days (range, 1–89 days).

MR spectroscopic imaging data were routinely obtained from the entire prostate of all patients. The total examination time of 25–30 minutes consisted of the following parts: 12–15 minutes for all anatomic T2-weighted imaging (transverse, sagittal, and coronal), 4 minutes for shimming, and 9 minutes for spectroscopic imaging measurements.

MR Imaging Acquisition
MR imaging was performed with a 3-T whole body unit (Magnetom Trio; Siemens Medical Solutions), with an eight-element body-array coil (31 patients) and the total imaging matrix concept (Siemens) (14 patients) for signal reception. In the latter case, the user selected the appropriate one or two three-element coil arrays from the supine coil matrix (eight arrays of three coil elements on the patient table) and one or two three-element coil arrays from the body matrix coil placed on top of the patient. Peristalsis was suppressed with an intramuscular injection of 1 mg glucagon (Glucagen; Novo Nordisk, Bagsvaerd, Denmark) immediately before the start of the examination.

With the patient inside the imager, a quick series of three orthogonal gradient-echo images (field of view, 400 x 400 mm; matrix size, 128 x 256; echo time, 5 msec; section thickness, 10 mm) was obtained for localization, after which the prostate anatomy and surrounding tissues were depicted with T2-weighted fast spin-echo MR imaging in three planes. Radiofrequency power deposition was reduced by using hyperechoes: Instead of a train of 180° pulses, the spin echoes were obtained with a train of low-power pulses with modulated flip angles to produce a full spin echo at the effective echo time (9). The parameters for the T2-weighted images were as follows: repetition time msec/effective echo time msec, 3500–5000/124; field of view, 220 mm; matrix size, 512 x 512; 13–17 sections; section thickness, 4 mm; intersection gap, 0.4 mm; number of signals acquired, two; and acquisition time per plane, 4–5 minutes.

MR Spectroscopic Imaging
After automatic and, if necessary, additional manual shimming (optimization of the main magnetic field homogeneity) of the prostate volume, a 1H MR spectroscopic imaging PRESS pulse sequence was performed to acquire proton MR spectra from every voxel of a three-dimensional (3D) excited subvolume containing the prostate. Because the two protons of the citrate resonance at 2.60 ppm are a strongly coupled spin system, the spectral shape of this resonance depends on magnetic field strength and pulse sequence timing (10). For an optimal shape of the citrate resonance at a magnetic field strength of 3 T, we used an echo time of 145 msec, at which the signal of citrate appears with large positive in-phase inner lines (7). This echo time is long enough for possible residual lipid signals to decay but is still short enough to depict the metabolites of interest. Water and lipids were suppressed with outer volume saturation slabs (Fig 1) and two 12.6-msec dual-frequency selective Mescher-Garwood pulses with crusher gradients (11,12).


Figure 1A
View larger version (174K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1a: T2-weighted MR imaging (3500/124) and 1H MR spectroscopic data in 61-year-old patient with prostate cancer. (a–c) T2-weighted MR images in transverse, coronal, and sagittal orientation, respectively, show placement of PRESS box (box around prostate) and outer volume saturation slabs (crosshatched bands). (d) Enlarged transverse image with one partition of 3D interpolated 1H MR spectroscopic matrix of 16 x 16 x 16 voxels (outer box). Outer volume saturation slabs extend into PRESS-selected volume of interest (inner box).

 

Figure 1B
View larger version (207K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1b: T2-weighted MR imaging (3500/124) and 1H MR spectroscopic data in 61-year-old patient with prostate cancer. (a–c) T2-weighted MR images in transverse, coronal, and sagittal orientation, respectively, show placement of PRESS box (box around prostate) and outer volume saturation slabs (crosshatched bands). (d) Enlarged transverse image with one partition of 3D interpolated 1H MR spectroscopic matrix of 16 x 16 x 16 voxels (outer box). Outer volume saturation slabs extend into PRESS-selected volume of interest (inner box).

 

Figure 1C
View larger version (205K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1c: T2-weighted MR imaging (3500/124) and 1H MR spectroscopic data in 61-year-old patient with prostate cancer. (a–c) T2-weighted MR images in transverse, coronal, and sagittal orientation, respectively, show placement of PRESS box (box around prostate) and outer volume saturation slabs (crosshatched bands). (d) Enlarged transverse image with one partition of 3D interpolated 1H MR spectroscopic matrix of 16 x 16 x 16 voxels (outer box). Outer volume saturation slabs extend into PRESS-selected volume of interest (inner box).

 

Figure 1D
View larger version (141K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1d: T2-weighted MR imaging (3500/124) and 1H MR spectroscopic data in 61-year-old patient with prostate cancer. (a–c) T2-weighted MR images in transverse, coronal, and sagittal orientation, respectively, show placement of PRESS box (box around prostate) and outer volume saturation slabs (crosshatched bands). (d) Enlarged transverse image with one partition of 3D interpolated 1H MR spectroscopic matrix of 16 x 16 x 16 voxels (outer box). Outer volume saturation slabs extend into PRESS-selected volume of interest (inner box).

 
A combination of an elliptic weighted averaged k-space acquisition scheme, zero filling to the nearest power of two, and 3D filtering of the signal in k-space was used (13) to reduce intervoxel signal contamination. Imaging parameters were as follows: 750/145; acquisition bandwidth, 1250 Hz; and 512 spectral data points. Field of view, matrix size, and number of signals acquired were adjusted in every prostate to achieve a nominal voxel resolution (before apodization) of 7 x 7 x 7 mm within a total acquisition time of approximately 9 minutes. After apodization, a voxel could best be approximated as a sphere with a diameter of 12.5 mm and a volume of 1.0 cm3 (13). Before Fourier transformation, the signals from the individual coil elements of the body-array coils were zero-order phased on the basis of the first point of the free induction decay and were added with the amplitude of that point as a weight factor (5).

To investigate the appearance of the strongly coupled resonances of polyamines between 3.0 and 3.3 ppm, a two-dimensional MR spectroscopic PRESS measurement of a phantom with a buffered solution of citrate, creatine, choline, and spermine at physiologic concentrations (90.0, 12.0, 9.5, and 18.0 mmol/L, respectively) was performed with the same pulse sequence timing as that used for in vivo spectroscopic imaging (echo time, 145 msec; time from 90° to first 180° pulse, 25.0 msec [7]; field of view of 96 x 96 mm, matrix of 12 x 12, section thickness of 10 mm, nominal voxel resolution before apodization of 8 x 8 x 10 mm, and acquisition time of 1 minute 52 seconds).

Histopathologic Analysis
Prostatectomy specimens were fixed overnight (in 10% neutral buffered formaldehyde) and coated with India ink. The prostate was sliced at 4-mm intervals in a plane parallel to the transverse T2-weighted sequence. All slices were routinely embedded in paraffin. Tissue slices of 5 µm thickness were prepared and stained with hematoxylin-eosin. The presence and extent of cancer were outlined on the glass cover and marked on photographs of the corresponding slice by a genitourinary pathologist (C.A.H., 13 years of experience) who was blinded to imaging results. All prostatectomy specimens were assigned a stage according to the 2002 TNM classification (14).

Spectroscopic Imaging Voxel Classification
On the basis of histopathologic findings and T2-weighted images in three directions with the overlaid spectroscopic imaging voxel matrix, a radiologist (S.W.T.P.J.H., 3 years of experience) in consensus with an MR spectroscopist (T.W.J.S., 4 years of experience) assigned 1–4 voxels of the 1H MR spectroscopic imaging matrix to each of the following tissues: healthy peripheral zone, healthy central gland (consisting of combined transition zone and central zone), healthy periurethral zone, and prostate cancer. Both reviewers were blinded to 1H MR spectroscopic imaging spectra. The inclusion of the periurethral zone as tissue separate from the central gland enabled us to study this region as a possible confounder in identifying cancer. The ejaculatory ducts, connecting the seminal vesicles through the central zone of the prostate to the urethra, can have high intensity signals from the ejaculatory ducts around 3.2 ppm. The voxels were assigned to different tissues on the basis of anatomic landmarks in the prostate visible on T2-weighted images (enabling distinction between central gland, peripheral zone, and urethra), rather than on the basis of hyper- or hypointensity of the tissue itself. Matching between histopathologic slices and T2-weighted images was estimated to be within an accuracy of one image section thickness (4 mm) in the craniocaudal direction (15).

For each assigned prostate cancer voxel, the size of the corresponding cancer focus as determined by using histopathologic findings was classified by the radiologist (S.W.T.P.J.H.) and spectroscopist (T.W.J.S.) in consensus as larger than 2 1H MR spectroscopic imaging voxels, between 1 and 2 1H MR spectroscopic imaging voxels, or smaller than 1 1H MR spectroscopic imaging voxel. To acknowledge the difficulties in aligning MR images and histopathologic slices (16), we assigned a certainty score to each selected cancer voxel, noted as a score of 1, possibly inside the cancer; a score of 2, probably inside the cancer; and a score of 3, definitely inside the cancer.

Data Processing and Analysis
For evaluation and quantification of all individual spectra, a software package (PRISMA; University of Bremen, Bremen, Germany, and Siemens Medical Solutions) was used. This software package uses a basis set of metabolic time signals of choline, creatine, and citrate, which are simulated by using literature values of chemical shifts and coupling constants. Especially for the complicated shape of citrate, this is an advantage, because the whole spectral shape is used in the fit (7). The reported amplitude of the citrate resonance is amplitude fitted to the model time signal, which can differ from the integral of the citrate resonance in the spectrum because its shape comprises both positive and negative signal intensities. After a frequency shift and residual water removal, the complex fit to the spectroscopic data in the time domain was performed, which included baseline artifact handling by truncation and remodeling of the first five data points. For every fitted metabolite, a Cramer-Rao lower bound was calculated, and a visual inspection of the original spectrum together with the curve fit and residual plot was performed by the spectroscopist (T.W.J.S.). Voxels with a correct automatic choice of the resonances (ie, PRISMA-produced signal fits at the correct parts-per-million positions), without lipid signal contamination and severe baseline distortions and with minimal intensity in the residual plots, passed the visual inspection. From these voxels, the choline-plus-creatine–to-citrate (Cho + Cr/Cit) ratio was calculated by dividing the sum of the fitted amplitudes in the free induction decay of choline and creatine by the fitted amplitude of the citrate signal.

Because the pathologist (C.A.H.) indicated the Gleason score for every outlined cancer focus in the histopathologic slices, we were able to investigate a possible relation between the Gleason score and the Cho + Cr/Cit ratio for every classified cancer voxel.

Statistical Analysis
The area under the receiver operating characteristic (ROC) curve (Az) for discrimination between cancer and healthy tissue was calculated by using the classified voxels as independent values. ROC curves were calculated with cancer voxels that had a matching certainty score of 2 or 3 (probably or definitely inside cancer). Separate analyses were performed for peripheral zone, central gland, and periurethral cancer localization. Sensitivity and specificity at different threshold values for the different tissues were calculated.

P values for comparisons of different tissues were calculated with the Bonferroni multiple comparison test after a one-way analysis of variance; a P value of .05 or less was considered to indicate a significant difference. P values for differences in ROC curves were calculated with two-sided paired t tests. The Spearman rank correlation test was used to calculate r values. Statistical analyses were performed by using software (MedCalc, version 8.1.0.0, MedCalc Software, Mariakerke, Belgium; Prism, version 4.00, GraphPad Software, San Diego, Calif; SPSS, version 12.0.1, SPSS, Chicago, Ill).


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 ADVANCES IN KNOWLEDGE
 IMPLICATION FOR PATIENT CARE
 References
 
Phantom Measurement and in Vivo 1H MR Spectroscopic Imaging
In a representative 1H MR spectroscopic imaging data set accompanied by histopathologic findings of the corresponding resected prostate, signals of citrate, choline, and creatine were depicted throughout the entire prostate (Fig 2). In many spectra in all patients, the choline and creatine resonances could not be clearly separated: Polyamine signals were still present at the used pulse-sequence timing, as is illustrated with the spectrum of a voxel inside the phantom with a solution of relevant prostate metabolites (Fig 3). Although the line width of the signals in this spectrum is quite small (4 Hz), the signals from the strongly coupled protons of spermine overlap with both creatine and choline signals. Therefore, in all voxels of the in vivo measurements with larger line widths and possibly other polyamines present, the complete region from approximately 3.0 to 3.2 ppm, reflected in a fit of two resonances to this region, was summed and used as the numerator of the marker ratio for prostate cancer.


Figure 2
View larger version (58K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2: Histopathologic findings and corresponding 3D 1H MR spectroscopic data (750/145) of prostate in 64-year-old patient with prostate cancer (prostate-specific antigen level, 6.86 ng/mL; final Gleason score, 3 + 4; stage, pT2c). Top: Photographs of histopathologic findings from apex to midgland in transverse slices. Arrows = areas of specimen that correspond to location of spectra and images below. Middle: Spectra from in vivo prostate. Left to right: healthy peripheral zone, cancer in peripheral zone, cancer in central gland, and healthy central gland. Scale is adjusted for each spectrum. Bottom: Transverse T2-weighted MR images of three of 16 sections of 3D 1H MR spectroscopic data from apex to midgland overlaid with corresponding spectral maps (range, 2.0–3.5 ppm). True size and location of voxels of which spectra are shown in middle row are indicated with circles. CP±? = possible capsular penetration, RV++ = positive resection margins.

 

Figure 3
View larger version (12K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 3: MR spectrum of voxel in phantom with buffered solution of citrate, creatine, choline, and spermine at physiologic concentrations (90.0, 12.0, 9.5, and 18.0 mmol/L, respectively). Spermine, one of the polyamines, has relatively broad relevant signal intensities in 3.0–3.2-ppm range at the used pulse timing. If line widths increase and other polyamines are also present, which is often the case in vivo, it is difficult to reliably quantify resonances of choline, creatine, and polyamines separately.

 
Voxel Classification and Metabolite Ratios
In total, 375 voxels were selected on the basis of histopathologic findings and assigned to one of the four tissues in all 45 patients: 87 voxels to cancer, 118 voxels to healthy peripheral zone, 115 voxels to healthy central gland, and 55 to healthy periurethral tissue. All selected voxels were processed with the automatic signal fit of the PRISMA software package (Fig 4). In four patients (26 voxels) (Fig 5), none of the classified voxels contained useful spectra. With the exclusion of these four patients, the average percentage per patient of voxels that passed the visual inspection of the automated fit procedure was 74% (median, 82%).


Figure 4A
View larger version (16K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 4a: Spectral fits and calculations of Cho + Cr/Cit (CC/C) ratio by using PRISMA software with data of three locations in prostate of 64-year-old patient with prostate cancer. Spectra originate from (a) healthy peripheral zone, (b) transition zone cancer with Gleason score of 2 + 4, and (c) peripheral zone cancer with Gleason score of 3 + 4. From top to bottom for each fit in (a–c) measured spectrum (black line) overlaid with fit (red line) and baseline (blue line), metabolite fit (green line), and residual between data and fit (black line).

 

Figure 4B
View larger version (15K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 4b: Spectral fits and calculations of Cho + Cr/Cit (CC/C) ratio by using PRISMA software with data of three locations in prostate of 64-year-old patient with prostate cancer. Spectra originate from (a) healthy peripheral zone, (b) transition zone cancer with Gleason score of 2 + 4, and (c) peripheral zone cancer with Gleason score of 3 + 4. From top to bottom for each fit in (a–c) measured spectrum (black line) overlaid with fit (red line) and baseline (blue line), metabolite fit (green line), and residual between data and fit (black line).

 

Figure 4C
View larger version (17K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 4c: Spectral fits and calculations of Cho + Cr/Cit (CC/C) ratio by using PRISMA software with data of three locations in prostate of 64-year-old patient with prostate cancer. Spectra originate from (a) healthy peripheral zone, (b) transition zone cancer with Gleason score of 2 + 4, and (c) peripheral zone cancer with Gleason score of 3 + 4. From top to bottom for each fit in (a–c) measured spectrum (black line) overlaid with fit (red line) and baseline (blue line), metabolite fit (green line), and residual between data and fit (black line).

 

Figure 5
View larger version (5K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 5: Flowchart of study profile. PCa = prostate cancer.

 
From 257 included spectra, both mean and median Cho + Cr/Cit ratios of the voxels from healthy tissues significantly differed from those of cancer tissue (Fig 6) (P < .001). The mean Cho + Cr/Cit ratio in the peripheral zone was significantly different from that in central gland and periurethral tissue (P < .05), whereas there was no statistical difference between ratios in the central gland and periurethral tissue (P = .90). However, some voxels in the periurethral area had the highest Cho + Cr/Cit ratios among healthy tissues. Focusing on voxels from cancer tissue only and using the additional classifications concerning cancer size and matching certainty, we found an increase in mean and median Cho + Cr/Cit ratios with both increasing cancer size and matching certainty (Fig 7).


Figure 6
View larger version (10K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 6: Graph of Cho + Cr/Cit (CC/C) ratio of tissues in prostate. Mean values, indicated with a horizontal line, for Cho + Cr/Cit ratio are 0.22 ± 0.12 (standard deviation) for peripheral zone (PZ), 0.34 ± 0.14 for central gland (CG), 0.36 ± 0.20 for periurethral area (U), and 1.3 ± 3.7 for cancer tissue. For display purposes, 6 voxels with Cho + Cr/Cit ratio greater than 1.5 are shown as 1.5.

 

Figure 7A
View larger version (9K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 7a: Graphs of valid cancer voxels classified by size and matching certainty score. (a) Cho + Cr/Cit (CC/C) ratio is plotted against size: class A = larger than 2 1H MR spectroscopic voxels (approximately 2 cm3), class B = between 1 and 2 1H MR spectroscopic voxels (between 1 and 2 cm3), and class C = smaller than 1 1H MR spectroscopic voxel (approximately 1 cm3). (b) Relationship with matching certainty score is shown: class 1 = possibly inside cancer, class 2 = probably inside cancer, and class 3 = definitely inside cancer. For display purposes, Cho + Cr/Cit ratios greater than 1.5 are shown as 1.5.

 

Figure 7B
View larger version (9K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 7b: Graphs of valid cancer voxels classified by size and matching certainty score. (a) Cho + Cr/Cit (CC/C) ratio is plotted against size: class A = larger than 2 1H MR spectroscopic voxels (approximately 2 cm3), class B = between 1 and 2 1H MR spectroscopic voxels (between 1 and 2 cm3), and class C = smaller than 1 1H MR spectroscopic voxel (approximately 1 cm3). (b) Relationship with matching certainty score is shown: class 1 = possibly inside cancer, class 2 = probably inside cancer, and class 3 = definitely inside cancer. For display purposes, Cho + Cr/Cit ratios greater than 1.5 are shown as 1.5.

 
ROC curves were calculated for tissues by using only cancer voxels that were probably or definitely inside cancer (matching certainty scores of 2 and 3), thereby excluding the category possibly inside cancer. In seven patients, only cancer voxels with a matching certainty score of 1 were classified, so ROC curves were constructed with cancer voxels from 34 patients and voxels in healthy tissues from 41 of 45 patients. Because the Cho + Cr/Cit distribution between the central gland and the periurethral area did not differ significantly, we combined the Cho + Cr/Cit distributions from the central gland and the periurethral area into a single ROC curve (Fig 8). The Az for the tissues were 0.84 (95% confidence interval: 0.76, 0.90) for the peripheral zone and 0.69 (95% confidence interval: 0.62, 0.76) for the combined central gland and periurethral area (P = .002). In the Table, the sensitivity and specificity at two threshold values for the Cho + Cr/Cit ratio for peripheral zone and central zone tissues are summarized.


Figure 8
View larger version (14K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 8: ROC curves summarize accuracy of discrimination of cancer from healthy peripheral zone and central gland tissue at 1H MR spectroscopy with only external surface coils at 3 T.

 

View this table:
[in this window]
[in a new window]

  		Sensitivity and Specificity of 1H MR Spectroscopy for Discrimination of Cancer from Healthy Peripheral Zone and Central Gland Tissue at Cho + Cr/Cit Thresholds

 
Although we used only the Cho + Cr/Cit ratio of those voxels that were classified as definitely inside cancer (class 3 only) to relate to cancer aggressiveness (Fig 9), we did not find a correlation between the Cho + Cr/Cit ratio and Gleason score (r = 0.17).


Figure 9
View larger version (6K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 9: Graph shows that correlation between Cho + Cr/Cit (CC/C) ratio of cancer voxels assigned with high matching certainty and total Gleason score of corresponding cancer focus at histopathologic examination was not found (r = 0.17).

 

   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
ADVANCES IN KNOWLEDGE
 IMPLICATION FOR PATIENT CARE
 References
 
Our study results demonstrate that 1H MR spectroscopic imaging of the prostate with the use of external surface coils at 3 T is feasible. The examinations—including MR imaging—were noninvasive and relatively short (<30 minutes). The number of classified voxels that passed visual inspection of the quality of the fit and that were chosen at random in tissues throughout the prostate of every patient can be used as an estimate of what percentage of the voxels of the data set were useful. With the exclusion of four patients because of inadequate shimming values and/or low signal-to-noise ratio, the median amount of 82% of useful voxels per patient reflects the robustness of the pulse sequence, the combination of signals from multiple coil elements, and the automatic fit routine. From these useful voxels, the Cho + Cr/Cit ratio could be calculated, which resulted in a Cho + Cr/Cit threshold value of 0.41. At 0.41, the sensitivity and specificity, respectively, of discrimination of cancer from healthy tissue were 66% and 95% for the peripheral zone and 66% and 73% for the central gland.

Although we used a different repetition time, echo time, and magnetic field strength, the mean and standard deviation of the Cho + Cr/Cit ratio of voxels in the healthy peripheral zone (0.22 ± 0.12) were almost identical to values reported previously with similar combined water and lipid suppression at 1.5 T with an endorectal coil (0.22 ± 0.13) (17). Apparently, differences in pulse sequence timing equalize with differences in metabolite relaxation times. On one hand, choline and creatine signals are lower than citrate signal because of a slight increase in saturation of these signals at 3 T (T1 of choline in the prostate at 3 T has been reported to be 1.6 seconds ± 0.5 compared with 0.47 second ± 0.14 for citrate) (7). On the other hand, this is compensated by the smaller loss of signal at an echo time of 145 msec because T2 relaxation for choline (T2 of 0.24 second ± 0.09) (7) is higher than that of citrate (T2 of 0.17 second ± 0.05) (7) at 3 T.

With complete coverage of the whole prostate at 3D 1H MR spectroscopic imaging, we could differentiate Cho + Cr/Cit values of the peripheral zone from values of the central gland. Another study (18) at 1.5 T on differentiation between central gland and peripheral zone at two-dimensional 1H MR spectroscopic imaging, with use of an endorectal coil and somewhat different MR parameters, revealed slightly deviating mean Cho + Cr/Cit values of healthy tissues (0.38 ± 0.15 for peripheral zone and 0.43 ± 0.16 for central gland) but a similar overlap of Cho + Cr/Cit values in cancer with those in healthy tissue. Mean Cho + Cr/Cit values for voxels in cancer tissue are difficult to compare between studies because the distribution in Cho + Cr/Cit values is far from Gaussian: A single or small number of very high Cho + Cr/Cit values dominate the mean value.

Because the mean Cho + Cr/Cit ratio of the central gland is higher than that of the peripheral zone, the overlap with voxels from cancer tissue is larger and the power of using the ratio to discriminate healthy tissue from cancer tissue is smaller, as is reflected in the lower Az of the ROC curve for the central gland than that for the peripheral zone. For a separate analysis of cancer in the central gland and the peripheral zone, the amount of matched voxels of cancer tissue in the central gland was too small. Therefore, voxels of cancer tissue were not subdivided according to their location in the prostate in the analysis.

The shape of the citrate signal at 3 T at the applied pulse timing calls for a postprocessing method that calculates more than the integral of signals in the spectrum only. The integral of the citrate signal varies with its line width and can approach zero when positive and negative lobes of the phased signal cancel each other. This problem was overcome by fitting a complex signal amplitude to a model function in the time domain. As mentioned earlier, the increase in spectral resolution obtained at 3 T compared with that at 1.5 T did not necessarily increase the accuracy of fitting choline and creatine resonances separately, because polyamine resonances between these two metabolites often still appeared to be present. Adding the integrals of these metabolites, thereby quantifying the complete group of resonances from 3.0 to 3.2 ppm and dividing them by the integral of citrate (represented by the fitted amplitude of the time domain signal), which is a general approach with spectra at 1.5 T, provided a more robust measure for quantifying tissues. A general problem with this approach remains partial cancellation of the sum of resonances by a decrease in polyamine signals together with an increase in choline signals in cancer tissue.

Difficulties in matching histopathologic slices with T2-weigthed MR images were acknowledged with the additional classification we made for the cancer voxels: Both smaller foci and foci that were matched with lower confidence had smaller Cho + Cr/Cit ratios. Furthermore, as a result of apodization to reduce voxel bleed (13), the true size of a voxel (approximately 1 cm3) was larger compared with what is called a clinically relevant cancer focus volume (>0.5 cm3) (19,20) and also was larger than voxel sizes reported at 1.5 T (0.24–0.7 cm3 [1] and 0.6–0.8 cm3 [18], both without apodization). This introduces partial volume effects. In smaller cancers, voxels could contain both cancer and healthy tissue with a corresponding lower Cho + Cr/Cit ratio, which is also reflected in the decrease in median Cho + Cr/Cit ratio with decreasing cancer size. Because the highest Cho + Cr/Cit values of healthy tissues were present around the urethra, we cannot rule out the urethra and ejaculatory ducts as possible confounders for cancer. However, the size of the voxels and the corresponding partial volume effects when defining a voxel around the urethra could be a reason why we did not find significant differences between the Cho + Cr/Cit ratio of the periurethral area and that of the healthy central gland. Throughout the central gland, including the urethra, the Cho + Cr/Cit ratio is generally larger than that of the healthy peripheral zone. Benign diseases of the central gland that increase the signals from 3.0 to 3.2 ppm or anatomic differences, such as less ductal tissue in which citrate is accumulated, could be reasons for this.

Although such findings were revealed in prostate cancer studies at 1.5 T (15), we were not able to detect a trend between the Cho + Cr/Cit ratio of a voxel in cancer tissue and the corresponding Gleason score at histopathologic analysis. A possible explanation could be that our patients were eligible for prostatectomy and therefore did not have the highest Gleason scores (predominantly 5–7). Consequently, we could not cover the complete range of Gleason scores proportionally with our patient population.

Our study had limitations. Without an endorectal coil, the signal-to-noise ratio inside the prostate was limited, which forced us to use a voxel size of 1.0 cm3, with its corresponding partial volume effects. If an endorectal coil had been used, the signal-to-noise ratio in the peripheral zone of the prostate would have been four to 10 times higher, depending on the actual distance from the voxel to the coil conductors. With increasing distance to the coil conductors (in central gland tissues), the benefit of an endorectal coil is less (21).

Not all voxels of all prostates were evaluated in our study. Although the classified voxels were chosen at random locations within the tissues of the prostate, they are a subset of the total number of voxels. But, because the classification was performed on the basis of histopathologic and MR imaging findings, with reviewers blinded to MR spectroscopic results and with the incorporation of both large and smaller cancer foci, we do not expect a bias in quantification with this subset. Furthermore, the MR spectroscopic findings were studied without an extensive comparison with sensitivity and specificity for distinguishing benign from malignant tissue with T2-weighted MR imaging alone at 3 T. Therefore, we are unable to estimate the incremental value of MR spectroscopic imaging over T2-weighted MR imaging—which has been found for 1.5 T (22)—for a field strength of 3 T.

Another limitation of our study was its relatively small number of patients. We did not assign a prospective score, or radiologic certainty score, to tissue for the presence of cancer, as has been proposed by Jung et al (23). We also did not take into account possibly elevated choline or decreased polyamine levels as separate parameters, which can have additional value. We investigated the quantitative value of the Cho + Cr/Cit ratio itself, without the experience of radiologic input but also without the possible bias of radiologic input, as a basis for either radiologic assignment or pure quantitative assessment of future prospective work.

In summary, on the basis of only the numbers of the Cho + Cr/Cit ratio, which originated from 1.0-cm3 voxels and was measured in less than 10 minutes, 3D proton MR spectroscopy at 3 T with only external array coils had a high specificity for the discrimination of cancer from healthy tissues; therefore, this technique shows potential for the metabolic identification of prostate cancer. Further and larger studies are needed, however, for confirmation.


   ADVANCES IN KNOWLEDGE
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 ADVANCES IN KNOWLEDGE
IMPLICATION FOR PATIENT CARE
 References
 


   IMPLICATION FOR PATIENT CARE
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 ADVANCES IN KNOWLEDGE
 IMPLICATION FOR PATIENT CARE
References
 


   ACKNOWLEDGMENTS
 
Parts of the protocol of our study have been derived from the ongoing International Multi-centre Assessment of Prostate MR Spectroscopy (IMAPS) study (24).


   FOOTNOTES
 

Abbreviations: Az = area under the ROC curve • Cho + Cr/Cit = choline plus creatine to citrate • PRESS = point-resolved spatially localized spectroscopy • ROC = receiver operating characteristic • 3D = three-dimensional

See Materials and Methods for pertinent disclosures.

Author contributions: Guarantors of integrity of entire study, T.W.J.S., A.H.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; manuscript final version approval, all authors; literature research, T.W.J.S., S.W.T.P.J.H., B.C.K., J.A.W., J.O.B., A.H.; experimental studies, T.W.J.S., S.W.T.P.J.H.; clinical studies, T.W.J.S., S.W.T.P.J.H., C.A.H., B.C.K., J.A.W., J.O.B.; statistical analysis, T.W.J.S., S.W.T.P.J.H.; and manuscript editing, all authors


   References
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 ADVANCES IN KNOWLEDGE
 IMPLICATION FOR PATIENT CARE
 References
 

  1. Kurhanewicz J, Vigneron DB, Hricak H, Narayan P, Carroll P, Nelson SJ. Three-dimensional H-1 MR spectroscopic imaging of the in situ human prostate with high (0.24-0.7-cm3) spatial resolution. Radiology 1996;198:795–805.[Abstract/Free Full Text]
  2. Heerschap A, Jager GJ, van der Graaf M, et al. In vivo proton MR spectroscopy reveals altered metabolite content in malignant prostate tissue. Anticancer Res 1997;17:1455–1460.[Medline]
  3. Kurhanewicz J, Vigneron DB, Nelson SJ, et al. Citrate as an in vivo marker to discriminate prostate cancer from benign prostatic hyperplasia and normal prostate peripheral zone: detection via localized proton spectroscopy. Urology 1995;45:459–466.[CrossRef][Medline]
  4. Lichy MP, Pintaske J, Kottke R, et al. 3D proton MR spectroscopic imaging of prostate cancer using a standard spine coil at 1.5 T in clinical routine: a feasibility study. Eur Radiol 2005;15:653–660.[CrossRef][Medline]
  5. Brown MA. Time-domain combination of MR spectroscopy data acquired using phased-array coils. Magn Reson Med 2004;52:1207–1213.[CrossRef][Medline]
  6. Bottomley PA. Spatial localization in NMR spectroscopy in vivo. Ann N Y Acad Sci 1987;508:333–348.[Medline]
  7. Scheenen TW, Gambarota G, Weiland E, et al. Optimal timing for in vivo (1)H-MR spectroscopic imaging of the human prostate at 3T. Magn Reson Med 2005;53:1268–1274.[CrossRef][Medline]
  8. van der Graaf M, Schipper RG, Oosterhof GO, Schalken JA, Verhofstad AA, Heerschap A. Proton MR spectroscopy of prostatic tissue focused on the detection of spermine, a possible biomarker of malignant behavior in prostate cancer. Magma 2000;10:153–159.[Medline]
  9. Hennig J, Scheffler K. Hyperechoes. Magn Reson Med 2001;46:6–12.[CrossRef][Medline]
  10. Wilman AH, Allen PS. The response of the strongly coupled AB system of citrate to typical 1H MRS localization sequences. J Magn Reson B 1995;107:25–33.[CrossRef][Medline]
  11. Mescher M, Tannus A, Johnson MO, Garwood M. Solvent suppression using selective echo dephasing. J Magn Reson A 1996;123:226–229.[CrossRef]
  12. Star-Lack J, Nelson SJ, Kurhanewicz J, Huang LR, Vigneron DB. Improved water and lipid suppression for 3D PRESS CSI using RF band selective inversion with gradient dephasing (BASING). Magn Reson Med 1997;38:311–321.[Medline]
  13. Scheenen TW, Klomp DW, Roll SA, Futterer JJ, Barentsz JO, Heerschap A. Fast acquisition-weighted three-dimensional proton MR spectroscopic imaging of the human prostate. Magn Reson Med 2004;52:80–88.[CrossRef][Medline]
  14. Greene FL, Page DL, Fleming ID, et al. AJCC cancer staging manual. 6th ed. New York, NY: Springer-Verlag, 2002.
  15. Zakian KL, Sircar K, Hricak H, et al. Correlation of proton MR spectroscopic imaging with Gleason score based on step-section pathologic analysis after radical prostatectomy. Radiology 2005;234:804–814.[Abstract/Free Full Text]
  16. Jager GJ, Ruijter ET, van de Kaa CA, et al. Local staging of prostate cancer with endorectal MR imaging: correlation with histopathology. AJR Am J Roentgenol 1996;166:845–852.[Abstract/Free Full Text]
  17. Males RG, Vigneron DB, Star-Lack O, et al. Clinical application of BASING and spectral/spatial water and lipid suppression pulses for prostate cancer staging and localization by in vivo 3D 1H magnetic resonance spectroscopic imaging. Magn Reson Med 2000;43:17–22.[CrossRef][Medline]
  18. van Dorsten FA, van der Graaf M, Engelbrecht MR, et al. Combined quantitative dynamic contrast-enhanced MR imaging and (1)H MR spectroscopic imaging of human prostate cancer. J Magn Reson Imaging 2004;20:279–287.[CrossRef][Medline]
  19. Augustin H, Hammerer PG, Graefen M, et al. Insignificant prostate cancer in radical prostatectomy specimen: time trends and preoperative prediction. Eur Urol 2003;43:455–460.[CrossRef][Medline]
  20. Stamey TA, Freiha FS, McNeal JE, Redwine EA, Whittemore AS, Schmid HP. Localized prostate cancer: relationship of tumor volume to clinical significance for treatment of prostate cancer. Cancer 1993;71:933–938.[CrossRef][Medline]
  21. Futterer JJ, Scheenen TW, Huisman HJ, et al. Initial experience of 3 Tesla endorectal coil magnetic resonance imaging and 1H-spectroscopic imaging of the prostate. Invest Radiol 2004;39:671–680.[CrossRef][Medline]
  22. Futterer JJ, Heijmink SW, Scheenen TW, et al. Prostate cancer localization with dynamic contrast-enhanced MR imaging and proton MR spectroscopic imaging. Radiology 2006;241:449–458.[Abstract/Free Full Text]
  23. Jung JA, Coakley FV, Vigneron DB, et al. Prostate depiction at endorectal MR spectroscopic imaging: investigation of a standardized evaluation system. Radiology 2004;233:701–708.[Abstract/Free Full Text]
  24. Scheenen T, van Hecke P, Bachert P, et al. First results of IMAPS: an international multi-centre assessment of prostate MR spectroscopy [abstr]. In: Radiological Society of North America scientific assembly and annual meeting program. Oak Brook, Ill: Radiological Society of North America, 2005; 419.



This article has been cited by other articles:


Home page
 RadioGraphicsHome page
K. J. Chang, I. R. Kamel, K. J. Macura, and D. A. Bluemke
3.0-T MR Imaging of the Abdomen: Comparison with 1.5 T
RadioGraphics, November 1, 2008; 28(7): 1983 - 1998.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow All Versions of this Article:
2451061444v1
245/2/507    most recent
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Scheenen, T. W. J.
Right arrow Articles by Heerschap, A.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Scheenen, T. W. J.
Right arrow Articles by Heerschap, A.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
RADIOLOGY RADIOGRAPHICS RSNA JOURNALS ONLINE