Combined T2-Weighted and Diffusion-Weighted MRI for Localization of Prostate Cancer

Transcription

1 T2 Combined with DWI for Prostate Cancer Localization Genitourinary Imaging Original Research Masoom A. Haider 1 Theodorus H. van der Kwast 2,3 Jeff Tanguay 2 Andrew J. Evans 2 Ali-Tahir Hashmi 1 Gina Lockwood 4 John Trachtenberg 5 Haider MA, van der Kwast TH, Tanguay J, et al. Keywords: diffusion-weighted imaging, MRI technique, oncologic imaging, prostate cancer, radiologic pathologic correlation, T2-weighted imaging DOI: /AJR Received September 28, 2006; accepted after revision March 28, Supported by grants from the Prostate Cancer Research Foundation of Canada and Princess Margaret Hospital Foundation. 1 Joint Department of Medical Imaging, Princess Margaret Hospital, University Health Network and Mount Sinai Hospital, University of Toronto, 610 University Ave., Toronto, ON M5G 2M9, Canada. Address correspondence to M. A. Haider. 2 Department of Pathology and Laboratory Medicine, Princess Margaret Hospital, University Health Network, University of Toronto, ON, Canada. 3 Department of Pathology and Laboratory Medicine, Mount Sinai Hospital, University of Toronto, ON, Canada. 4 Department of Biostatistics, Princess Margaret Hospital, University Health Network, University of Toronto, ON, Canada. 5 Department of Surgical Oncology, Princess Margaret Hospital, University Health Network, University of Toronto, ON, Canada. AJR 2007; 189: X/07/ American Roentgen Ray Society Combined T2-Weighted and Diffusion-Weighted MRI for Localization of Prostate Cancer OBJECTIVE. The objective of our study was to compare T2-weighted MRI alone and T2 combined with diffusion-weighted imaging (DWI) for the localization of prostate cancer. SUBJECTS AND METHODS. T2-weighted imaging and DWI (b value = 600 s/mm 2 ) were performed in 49 patients before radical prostatectomy using an endorectal coil at 1.5 T in this prospective trial. The peripheral zone of the prostate was divided into sextants and the transition zone into left and right halves. T2 images alone and then T2 images combined with apparent diffusion coefficient (ADC) maps (T2 + DWI) were scored for the likelihood of tumor and were compared with whole-mount histology results. Fixed window and level settings were used to display the ADC maps. Only tumors with an area of more than 0.13 cm 2 (> 4 mm diameter) and a Gleason score of 6 were considered significant. The area under the receiver operating characteristic curve (A z ) was used to assess accuracy. RESULTS. In the peripheral zone, the A z value was significantly higher (p = 0.004) for T2 plus DWI (A z = 0.89) than for T2 imaging alone (A z = 0.81). Performance was poorer in the transition zone for both T2 plus DWI (A z = 0.78) and T2 (A z = 0.79). For the whole prostate, sensitivity was significantly higher (p < 0.001) with T2 plus DWI (81% [120/149]) than with T2 imaging alone (54% [81/149]), with T2 plus DWI showing only a slight loss in specificity compared with T2 imaging alone (84% [204/243] vs 91% [222/243], respectively). CONCLUSION. Combined T2 and DWI MRI is better than T2 imaging alone in the detection of significant cancer (Gleason score 6 and diameter > 4 mm) within the peripheral zone of the prostate. ocalization of prostate cancer is important given the emergence of dis- L ease-targeted therapies, such as intensity-modulated radiation therapy, interstitial brachytherapy, and cryosurgery, as part of patient care [1]. Knowledge of the tumor location within the prostate can help direct maximal therapy to the largest focus of tumor while minimizing damage to the surrounding structures, such as the neurovascular bundles, the rectal wall, and the neck of the bladder. Studies have shown the added value of T2- weighted MRI and proton spectroscopy (MR spectroscopy) in localizing prostate cancer compared with endorectal sonographically guided biopsy and digital rectal examination, which are the traditional methods of determining prostate cancer location [2, 3]. T2-weighted imaging has been shown to provide some localization information in this setting, with previously reported sensitivities and specificities of 67 81% and 46 69%, respectively. MR spectroscopy has shown promise in prostate cancer localization with a sensitivity of 73% and a specificity of 80%; however, in these studies up to 26% of the sextants were inadequately evaluated by spectroscopy and were eliminated from the analysis [3, 4]. This has motivated investigation into other MRI methods for prostate cancer detection. Diffusion-weighted imaging (DWI) is an MRI method typically used in neuroradiology. From DWI parametric maps, apparent diffusion coefficients (ADCs) can be calculated. Recently, a number of investigators have reported the potential usefulness of DWI for detecting prostate cancer because it shows a lower ADC than a normal peripheral zone [2, 5 7]. The purpose of this study was to compare the accuracy of T2-weighted MRI alone and T2 combined with DWI for the localization of prostate cancer. Subjects and Methods Patients Informed consent was obtained from all patients, and approval for this study was obtained AJR:189, August

2 TABLE 1: Characteristics of 49 Study Patients with Prostate Cancer and Tumors Characteristic Value Patient age (y) Median 61 Range PSA level (ng/ml) Median Range Pathologic stage T2a 7 T2b 0 T2c 26 T3a 13 T3b 1 T4 2 Maximum Gleason score No. (%) of locations with Gleason score of 6 246/392 (63) Area a of > 0.13 cm 2 and 149/392 (38) Gleason score of 6 Peripheral zone 127/149 (85) Base 39/127 (31) Mid 51/127 (40) Apex 37/127 (29) Transition zone 22/149 (15) Total locations 392 Note PSA = prostate-specific antigen. a With 4 mm diameter. from the ethics review board of our institution. Fifty-one men undergoing staging MRI before radical prostatectomy for prostate carcinoma were enrolled as part of this prospective trial from May 18, 2005, to May 5, One patient was excluded because of image distortion caused by a hip prosthesis, and another was excluded because the image acquisition had failed, thus leaving 49 patients in the study group for analysis (Table 1). The median time between MRI and prostatectomy was 15 days (range, 2 77 days). There was a minimum of 6 weeks between prostate biopsy and MRI. All patients were free of nodal and bone metastases on the basis of preoperative imaging findings. MRI Protocol Axial and coronal T2-weighted images and axial diffusion-weighted images were obtained. All examinations were performed using a 1.5-T MRI system (EchoSpeed, Excite, or Excite HD; GE Healthcare) and a 4-channel phased-array surface coil coupled to an endorectal coil. Imaging parameters for the T2 images were TR range/te, 4,800 5,650/96; echo-train length, 16; bandwidth, khz; field of view, 14 cm; slice thickness, 3 mm; gap, 0 mm; number of excitations, 3; no phase wrap; phase-encoding direction, left to right; and matrix, Next, an axial echo-planar DWI pulse sequence (b value = 600 s/mm 2 ) with the same slice locations as the T2 sequence was performed using the following parameters: TR/TE, 4,000/73; bandwidth, 167 khz; field of view, 14 cm; slice thickness, 3 mm; number of excitations, 8; matrix, ; 3 acquisitions with 8 slices per acquisition; ramp sampling; and optimized TE options. The time required to acquire the DWI image set was 8 minutes. Image Analysis and Interpretation ADC maps were generated using in-house software developed using IDL (version 6.0, ITT Visual Information Solutions). These maps were converted to 8-bit gray-scale images with a window width of 1, mm 2 /s and a window level of 1, mm 2 /s and were displayed using ImageJ software (U.S. National Institutes of Health). With this method, regions of low ADC were darker than regions of high ADC, with areas less than or equal to mm 2 /s being black. By using a fixed window width and level, we were able to take advantage of the quantitative nature of ADC mapping. T2-weighted images were reviewed on a workstation (efilm, version 2.1, Merge emed). The peripheral zone of the prostate was divided into base, mid, and apex and left and right halves, thus yielding six regions. The central gland comprising the transition zone was divided into left and right halves. The base was defined as the region extending from the most superior margin of the prostate to the widest transverse diameter of the prostate. The mid gland was defined as the region between the widest transverse diameter and the orifices of the ejaculatory ducts at the verumontanum. The apex was defined as the region inferior to the mid gland. All images were reviewed by a single radiologist with 11 years of experience interpreting body MRI and 6 years of experience interpreting prostate MRI. The T2 images were reviewed first. The ADC maps were subsequently reviewed in conjunction with the T2 images. The observer assigned a score to each zone using the following 5-point scale: 0, definitely no cancer; 1, probably no cancer; 2, possible cancer; 3, probable cancer; and 4, definite cancer. The criteria for these scores in the peripheral zone and in the transition zone of the prostate are listed in Table 2. The observer was blinded to the pathology results. Pathology After fixation in buffered formalin, prostatectomy specimens were inked with three different colors of stain to identify the left, right, and anterior sides of the specimen. After removal of the apex and the bladder neck resection margins, the prostate was sectioned axially to match the MRI plane of section at regular intervals of 4 mm or less, yielding serial slices of tissue. A ruler was used to ensure no single slice was more than 4 mm in thickness. These slices were sectioned into two halves (left and right) or four quarters (anterior and posterior, left and right) to fit on a standard slide. On each slide, a pathologist outlined the region of cancer and assigned a Gleason score. The slides were then digitized and reconstructed into whole-mount sections using Photoshop (Adobe Systems) at 300 dpi. From these digitized slides, each tumor region was traced by the pathologist to obtain a cross-sectional area using ImageJ software. The pathologist was blinded to the MRI results. For the purposes of radiologic pathologic correlation, a region was considered positive for cancer if it contained tumor with a cross-sectional area that was greater than 0.1 cm 2 on the fixed specimen with a Gleason score of 6 or higher. Assuming an ellipsoid shape of the tumor and accounting for tissue shrinkage after fixation and processing (linear factor = 1.14) [8], this would be equivalent to a geometric mean diameter of 4 mm and an area of 0.13 cm 2. In cases in which a sextant contained more than one tumor, the largest one was used. The radiologist reviewed the pathologic specimens in conjunction with the MR images to spatially match tumors in each zone. This was done 6 weeks after completion of MR image review. Each pathologic slice was visually matched to a corresponding MR image on the basis of the location of the ejaculatory ducts, diameter of the prostate, and approximate distance from the base or apex. To be considered a match, a tumor not only had to be in the same region from superior to inferior in the prostate but also had to lie in the same anterior or posterior half of the prostate. Statistical Analysis Receiver operating characteristic (ROC) curves were estimated separately for each region of the prostate using maximum likelihood and assuming bivariate normal data using ROCKIT software (version Beta, CE Metz, University of Chicago) [9]. The combined area under the ROC curve (A z ) was obtained by averaging the areas under the ROC 324 AJR:189, August 2007

3 T2 Combined with DWI for Prostate Cancer Localization TABLE 2: Scoring Scheme for Prostate Cancer Score T2-Weighted Imaging Alone T2 + Diffusion-Weighted Imaging Peripheral zone 0, Definitely not cancer Normal signal ADC normal or T2 = 0 1, Probably not cancer Mild low signal and diffuse or linear (feathered) appearance ADC borderline and T2 2 2, Possible cancer Mild low signal, focal but not clearly masslike or moderate signal drop and ADC borderline and T2 = 3 diffuse distribution 3, Probably cancer Mild signal drop and masslike appearance ADC abnormal, size < 3 mm, and T2 > 0 4, Definite cancer Moderate signal drop with a definite masslike appearance ADC abnormal, size > 3 mm, and T2 > 0 or T2 = 4 Transition zone 0, Definitely not cancer No nodules > 5 mm ADC normal or T2 = 0 1, Probably not cancer Internally heterogeneous nodules with well-defined borders and capsules ADC borderline and T2 2 2, Possible cancer Homogeneous low T2 intensity with preserved capsule ADC borderline and T2 3 or ADC abnormal and T2 2 3, Probably cancer Homogeneous low T2 signal with no capsule ADC abnormal and T2 = 3 4, Definite cancer Features of probable cancer and lenticular shape or invasion of anterior fibromuscular stroma or invasion into peripheral zone ADC abnormal and T2 = 4 Note ADC = apparent diffusion coefficient. True-Positive Fraction False-Positive Fraction curves for each region. This was done for the ratings of T2 imaging alone and for the combined T2 plus DWI ratings over all eight regions, just the peripheral zone (six regions), and just the transition zone (two regions). The jackknife method was used to estimate the variances of the A z derived from multiple regions of the prostate and to compare the correlated A z between the T2 imaging alone and T2 plus DWI assessments [10]. For visual display, the ROC curves obtained from the pooled data were plotted. Sensitivity, specificity, and positive and negative predictive values were calculated by choosing a threshold score of 3 or greater to indicate cancer. To account for any correlations due to multiple regions being assessed True-Positive Fraction False-Positive Fraction A B C Fig. 1 Receiver operator characteristic (ROC) curves for detection of prostate cancer on T2-weighted imaging alone (dashed line) and combined T2 and diffusion-weighted imaging (DWI) (solid line). A, For whole prostate, area under ROC curve (A z ) was significantly higher (p = 0.006) for T2 plus DWI (A z = 0.87) than for T2 imaging alone (A z =0.8). B, For peripheral zone, A z was significantly higher (p = 0.004) for T2 plus DWI (A z = 0.89) than for T2 imaging alone (A z =0.81). C, Test performance was poor for transition zone, with T2 imaging alone (A z = 0.79) and T2 plus DWI (A z = 0.78) showing similar results. for each patient, generalized estimating equation models were used to calculate the corresponding variances and 95% CIs and to test for differences between methods [11]. True-Positive Fraction False-Positive Fraction Results Of the 392 zones, 246 contained cancers with a Gleason score of 6 or more, 149 of which contained tumors with an area of more than 0.13 cm 2. Of the 97 zones with tumors 0.13 cm 2 or smaller, 91 (94%) had a Gleason score of 6. The A z value was significantly higher (p = 0.006) for T2 plus DWI (A z =0.87) than for T2 imaging alone (A z = 0.80) when evaluating the entire prostate. This was particularly true in the peripheral zone, where the A z value was 0.89 for T2 plus DWI and 0.81 for T2 (p = 0.004) (Figs. 1A and 1B). Performance was poorer in the transition zone for both T2 plus DWI (A z = 0.78) and T2 (A z =0.79) methods. T2 plus DWI did not improve tumor localization compared with T2 imaging alone in the transition zone (p = 0.90) (Fig. 1C). For cancers anywhere in the prostate, using a threshold score of 3 or greater to indicate cancer, sensitivity was significantly higher for T2 plus DWI than for T2 imaging alone without overlap of the CIs (81% [120/149]; 95% CI, 74 86% vs AJR:189, August

4 54% [81/149]; 95% CI, 44 65%, respectively; p < 0.001), and specificity was maintained above 80% (Figs. 2 and 3 and Table 3). The negative predictive value was also significantly higher for T2 plus DWI versus T2 imaging alone (88% [204/233]; 95% CI, 82 92% vs 77% [222/290]; 95% CI, %, respectively; p < 0.001) (Table 3). DWI added little to the detection of cancer in the transition zone with low sensitivities for T2 imaging alone (36% [8/22]; 95% CI, 16 63%) and for T2 plus DWI (41% [9/22]; 95% CI, 19 67%). A B C Fig. 2 Tumors seen on apparent diffusion coefficient (ADC) map and not on T2-weighted images in 46-year-old man with prostate tumor. Prostate-specific antigen was 9.9 ng/ml and Gleason score was 7. A, T2-weighted image shows generalized decrease in signal in peripheral zone and no focal mass on left side; this region was scored as 2, possible cancer. B, ADC map calculated from diffusion-weighted images at b values of 0 and 600 s/mm 2 shown at window width of 1, mm 2 /s and level of 1, mm 2 /s reveals clear focal mass in left peripheral zone (arrow); this region was scored as 4, definite cancer. C, Photomicrograph of pathologic specimen shows Gleason score of 7 (3 + 4) tumor (outlined areas). ED = ejaculatory duct. A B C Fig. 3 Definite prostate cancer seen on apparent diffusion coefficient (ADC) map but poorly seen on T2 images in 66-year-old man. Prostate-specific antigen was 3.2 ng/ml and Gleason score was 7 (3 + 4). A, T2-weighted image shows mild loss of signal at left and right posterolateral aspects of peripheral zone (arrows). These regions were scored as 2, possible cancer. B, ADC map calculated from diffusion-weighted images at b values of 0 and 600 s/mm 2 shown at window width of 1, mm 2 /s and level of 1, mm 2 /s shows focal area of decreased signal corresponding to one of two areas seen on T2 images in left posterolateral peripheral zone (arrow). This area was characterized as 4, definite cancer. C, Photomicrograph of pathologic specimen shows Gleason score of 7 (3 + 4) tumor (outlined areas). ED = ejaculatory duct. Discussion In this study, we have shown T2 combined with DWI was able to detect and localize prostate cancer better than T2-weighted imaging alone, primarily by providing increased sensitivity while maintaining high specificity in the peripheral zone. Prior studies have shown lower ADC values in prostate cancer than in a normal peripheral zone, but prospective whole-mount localization studies were limited at the time of the writing of this article. T2-weighted imaging has been shown to provide some localization information, with previously reported sensitivities and specificities for prostate cancer of 67 81% and 46 69% and an A z value of 0.72 [3, 4]. The sensitivity of T2-weighted imaging alone for the whole prostate was slightly lower at 54% and specificity was higher at 91% in our study, with a similar A z of This difference is likely related to the scoring scheme used, which required scoring a masslike appearance in the peripheral zone as a probable cancer and discounting heterogeneous nodules in the transition zone, which could be considered high-specificity review criteria. The mechanism by which DWI adds diagnostic accuracy to T2 imaging is uncertain. T2 signal loss in the peripheral zone may be related to a number of causes, including cancer, inflammation, fibrosis, and hemorrhage [12 14]. It is possible that the ADC value is more drastically altered by cancer than it is by factors such as hemorrhage, inflammation, or fibrosis. The cause of lower ADC values in prostate cancer may be related to the many tightly packed glandular elements found in cancers that locally replace the fluid-contain- 326 AJR:189, August 2007

5 T2 Combined with DWI for Prostate Cancer Localization TABLE 3: Sensitivity, Specificity, Predictive Values, and Accuracy of T2 Imaging Alone Versus Combined T2 and Diffusion- Weighted Imaging (DWI) for Prostate Cancer Performance Measure Sensitivity 81/149 (54) [44 65] Specificity 222/243 (91) [86 95] PPV 81/102 (79) [69 87] NPV 222/290 (77) [69 82] Accuracy 303/392 (77) [72 82] Whole Prostate Peripheral Zone Transition Zone T2 Alone T2 + DWI p T2 Alone T2 + DWI p T2 Alone T2 + DWI p 120/149 (81) [74 86] 204/243 (84) [77 89] 120/159 (75) [66 83] 204/233 (88) [82 92] 324/392 (83) [78 86] < /127 (57) [48 67] /167 (91) [85 95] /88 (83) [73 90] < /206 (74) [66 81] /294 (77) [71 82] 111/127 (87) [81 92] 134/167 (80) [72 87] 111/144 (77) [69 84] 134/150 (89) [83 93] 245/294 (83) [79 87] ing peripheral zone ducts. This could result in a local drop in the ADC value. The different natures of ADC and T2-weighted tissue contrast in the prostate are supported by at least one prior study showing no significant correlation between quantitative T2 measurement and ADC values in the prostate [15]. Another reason for the improved performance of DWI may be the quantitation of ADC, which theoretically eliminates the effect of T2 signal variations as well as receiver gain and coil intensity profiles from the image, thereby allowing fixed window levels for assessment. Prior studies have shown ADC values to be lower in prostate cancer, ranging from 1,100 to 1, mm 2 /s with b values of 300 1,000 mm 2 /s compared with normal peripheral zone values of 1,610 1, mm 2 /s [5, 6, 15]. The window and level settings chosen to display the ADC images in this study are consistent with settings that allow cancer to be distinguished from normal peripheral zone. Nonquantitative T2 imaging relies on visual assessment of relative signal changes between the tumor and normal peripheral zone and of the morphologic appearance of areas of low signal for the diagnosis, which may be more subjective and may affect test performance. Other methods of cancer detection in the prostate include MR spectroscopy and dynamic contrast-enhanced MRI, with many of the published articles in the literature focusing on staging. MR spectroscopy has shown promise in prostate cancer localization, with a sensitivity of 73% and a specificity of 80% [3, 4]; however, in those studies, up to 26% of the sextants were inadequately evaluated by spectroscopy and were eliminated from the analysis. Localization studies with dynamic contrast-enhanced MRI have shown a sensitivity and specificity of 73% and 77%, respectively, using a 3-T MRI system [16]. In contrast to these methods, DWI has the advantages of not requiring IV contrast material and of being simple to process. Moreover, DWI requires less time to acquire than proton spectroscopy and less technologist training to perform. DWI has some limitations. In this study, an echoplanar imaging based pulse sequence was used. This sequence can be affected by magnetic susceptibility, resulting in spatial distortion and signal loss. Alternative DWI methods based on line scan diffusion [17] or on the addition of parallel imaging can help reduce distortion and may help further improve diagnostic accuracy. MRI detection of prostate cancer is dependent on the size of the tumor. Tumor size is related to the risk of extracapsular spread [18], relapse after radical prostatectomy [19], and PSA progression [20]; and the larger the tumor, the higher the risk of treatment failure. Prior studies have shown that 89% of tumors less than 0.5 cm 3 (1 cm in diameter) are a Gleason score of 6 [21]. Such tumors are likely indolent [20, 22]. This suggests that the tumor size cutoff used in our study of 4 mm is sufficient to detect significant cancers in a large proportion of patients. This study has limitations. Interobserver variability was not assessed. Our standard T2 interpretation results are consistent with those of prior studies. A standardized interpretation scheme was also used for both the T2 imaging alone and the T2 plus DWI data sets (Table 2). We also used fixed window level settings for ADC map display, which should have added < /22 (36) [16 63] /76 (92) [80 97] 0.2 8/14 (57) [25 84] < /84 (83) [72 91] /98 (80) [69 88] 9/22 (41) [19 67] 70/76 (92) [80 97] 9/15 (60) [28 85] 70/83 (84) [73 91] 79/98 (81) [70 88] Note Regions scored as 3 or greater were considered positive (see Table 2). Boldface indicates statistically significant differences. Percentages are in parentheses and 95% CIs are in square brackets. PPV = positive predictive value, NPV = negative predictive value. < robustness to the interpretation, to take advantage of the quantitative nature of DWI. There was a bias toward patients with established cancer in this study because all patients had cancer proven on prior biopsy and were proceeding to radical prostatectomy. Thus, the population of patients with small tumors or with elevated PSA values resulting from inflammation or other causes may not have been fully represented. Finally, precise radiologic pathologic correlations are made difficult by tissue deformation, difficulty in precise matching of slicing angles to the plane of imaging, and deformation of the prostate at the time of MRI by the endorectal coil. We attempted to minimize errors in this study through careful radiologic pathologic correlation and matching size and internal structures such as the ejaculatory ducts. Our results suggest a potential role for DWI in providing localization before sonographically guided biopsy in patients with persistently elevated PSA values and prior negative biopsy results. Further investigation of this patient group is required to determine whether DWI can help improve the detection of prostate cancer in these patients. In conclusion, T2 plus DWI MRI is significantly better than T2-weighted imaging alone in the detection of significant cancer (Gleason score 6 and diameter > 4 mm) within the peripheral zone of the prostate. References 1. Carroll PR, Presti JC Jr, Small E, Roach M 3rd. Focal therapy for prostate cancer 1996: maximizing outcome. Urology 1997; 49: Mullerad M, Hricak H, Kuroiwa K, et al. Compar- AJR:189, August

6 ison of endorectal magnetic resonance imaging, guided prostate biopsy and digital rectal examination in the preoperative anatomical localization of prostate cancer. J Urol 2005; 174: Wefer AE, Hricak H, Vigneron DB, et al. Sextant localization of prostate cancer: comparison of sextant biopsy, magnetic resonance imaging and magnetic resonance spectroscopic imaging with step section histology. J Urol 2000; 164: Scheidler J, Hricak H, Vigneron DB, et al. Prostate cancer: localization with three-dimensional proton MR spectroscopic imaging clinicopathologic study. Radiology 1999; 213: Hosseinzadeh K, Schwarz SD. Endorectal diffusion-weighted imaging in prostate cancer to differentiate malignant and benign peripheral zone tissue. J Magn Reson Imaging 2004; 20: Issa B. In vivo measurement of the apparent diffusion coefficient in normal and malignant prostatic tissues using echo-planar imaging. J Magn Reson Imaging 2002; 16: Kozlowski P, Chang SD, Jones EC, Berean KW, Chen H, Goldenberg SL. Combined diffusionweighted and dynamic contrast-enhanced MRI for prostate cancer diagnosis: correlation with biopsy and histopathology. J Magn Reson Imaging 2006; 24: Schned AR, Wheeler KJ, Hodorowski CA, et al. Tissue-shrinkage correction factor in the calculation of prostate cancer volume. Am J Surg Pathol 1996; 20: Metz CE, Herman BA, Shen JH. Maximum likelihood estimation of receiver operating characteristic (ROC) curves from continuously-distributed data. Stat Med 1998; 17: Hajian-Tilaki KO, Hanley JA, Joseph L, Collet JP. Extension of receiver operating characteristic analysis to data concerning multiple signal detection tasks. Acad Radiol 1997; 4: Leisenring W, Pepe MS, Longton G. A marginal regression modelling framework for evaluating medical diagnostic tests. Stat Med 1997; 16: Shukla-Dave A, Hricak H, Eberhardt SC, et al. Chronic prostatitis: MR imaging and 1 H MR spectroscopic imaging findings initial observations. Radiology 2004; 231: Lovett K, Rifkin MD, McCue PA, Choi H. MR imaging characteristics of noncancerous lesions of the prostate. J Magn Reson Imaging 1992; 2: White S, Hricak H, Forstner R, et al. Prostate cancer: effect of postbiopsy hemorrhage on interpretation of MR images. Radiology 1995; 195: Gibbs P, Tozer DJ, Liney GP, Turnbull LW. Comparison of quantitative T2 mapping and diffusionweighted imaging in the normal and pathologic prostate. Magn Reson Med 2001; 46: Kim CK, Park BK, Kim B. Localization of prostate cancer using 3T MRI: comparison of T2-weighted and dynamic contrast-enhanced imaging. J Comput Assist Tomogr 2006; 30: Chan I, Wells W 3rd, Mulkern RV, et al. Detection of prostate cancer by integration of line-scan diffusion, T2-mapping and T2-weighted magnetic resonance imaging: a multichannel statistical classifier. Med Phys 2003; 30: Partin AW, Epstein JI, Cho KR, Gittelsohn AM, Walsh PC. Morphometric measurement of tumor volume and per cent of gland involvement as predictors of pathological stage in clinical stage B prostate cancer. J Urol 1989; 141: Epstein JI, Carmichael M, Partin AW, Walsh PC. Is tumor volume an independent predictor of progression following radical prostatectomy? A multivariate analysis of 185 clinical stage B adenocarcinomas of the prostate with 5 years of followup. J Urol 1993; 149: Kikuchi E, Scardino PT, Wheeler TM, Slawin KM, Ohori M. Is tumor volume an independent prognostic factor in clinically localized prostate cancer? J Urol 2004; 172: Olumi AF, Richie JP, Schultz DJ, D Amico AV. Calculated volume of prostate cancer identifies patients with clinical stage T1C disease at high risk of biochemical recurrence after radical prostatectomy: a preliminary study. Urology 2000; 56: Ohori M, Wheeler TM, Dunn JK, Stamey TA, Scardino PT. The pathological features and prognosis of prostate cancer detectable with current diagnostic tests. J Urol 1994; 152: AJR:189, August 2007