- Email: [email protected]
3T multiparametric MRI of the prostate: Does intravoxel incoherent motion diffusion imaging have a role in the detection and stratification of prostate cancer in the peripheral zone?
European Journal of Radiology 85 (2016) 790–794
Contents lists available at ScienceDirect
European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad
3T multiparametric MRI of the prostate: Does intravoxel incoherent motion diffusion imaging have a role in the detection and stratification of prostate cancer in the peripheral zone? Mariacristina Valerio a , Chiara Zini a , Davide Fierro a , Francesca Giura a , Anna Colarieti a , Alessandro Giuliani b , Andrea Laghi a , Carlo Catalano a , Valeria Panebianco a,∗ a b
Dept. of Radiological Sciences, Oncology & Pathology—Sapienza University of Rome, V.le Regina Elena, 324, 00161 Roma, Italy Environment and Health Dept., Istituto Superiore di Sanità, Rome, V.le Regina Elena, 299, 00161 Roma, Italy
a r t i c l e
i n f o
Article history: Received 30 October 2015 Received in revised form 13 January 2016 Accepted 16 January 2016 Keywords: Multiparametric magnetic resonance imaging Diffusion weighted imaging Intravoxel incoherent motion Prostate cancer Peripheral zone Gleason score
a b s t r a c t Purpose: To evaluate the potential added value of the intravoxel incoherent motion model to conventional multiparametric magnetic resonance protocol in order to differentiate between healthy and neoplastic prostate tissue in the peripheral zone. Material and methods: Mono-exponential and bi-exponential fits were used to calculate ADC and IVIM parameters in 53 patients with peripheral zone biopsy proved tumor. Inferential statistics analysis was performed on T2, ADC and IVIM parameters (D, D*, f) comparing healthy and neoplastic tissues. Linear discriminant analysis was performed for the conventional parameters (T2 and ADC), the IVIM parameters (molecular diffusion coefficient (D), perfusion-related diffusion coefficient (D*), and perfusion fraction (f) and the combined T2-weighted imaging/DWI and IVIM parameters (T2, ADC, D, D* and f). A correlation with Gleason scores was achieved. Results: The values of T2, ADC and D were significantly lower in cancerous tissues (2749.82 ± 1324.67 ms, 0.76 ± 0.27 × 10−3 mm2 /s and 0.99 ± 0.38 × 10−3 mm2 /s respectively) compared to those found in the healthy tissues (3750.70 ± 1735.37 ms, 1.39 ± 0.48 × 10−3 mm2 /s and 1.77 ± 0.36 × 10−3 mm2 /s respectively); D* parameter was significantly increased in neoplastic compared to healthy tissue (15.56 ± 12.91 × 10−3 mm2 /s and 10.25 ± 10.52 × 10−3 mm2 /s respectively). The specificity, sensitivity and accuracy of the T2-weighted imaging/DWI and IVIM parameters were 100, 96 and 98%, respectively, compare to 88, 92 and 90% and 96, 92 and 94 for T2-weighted imaging/ADC and IVIM alone. Conclusions: IVIM parameters increase the specificity and sensitivity in the evaluation of peripheral zone prostate cancer. A statistical difference between low grade tumors and high grade tumors has been demostrated in that ADC, D and D* dataset; in particular, D has been found to have the highest significativity. © 2016 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Multiparametric magnetic resonance (mp-MR) combined with magnetic resonance (MR)—targeted biopsy is considered valuable tool for early detection of clinically significant prostate cancer (PCa) [1]. In a recent randomized study comparing mp-MRI/biopsy vs. TRUS-guided biopsy, accuracy of mp-MRI in detecting clinically significant PCa was 97% [2].
∗ Corresponding author. E-mail addresses: [email protected], [email protected] (V. Panebianco). http://dx.doi.org/10.1016/j.ejrad.2016.01.006 0720-048X/© 2016 Elsevier Ireland Ltd. All rights reserved.
Among the various sequences composing the mp-MR acquisition protocol, diffusion-weighted imaging (DWI) has been claimed to contribute to T2-weighted MRI in the detection and localization of PCa; in particular, in the peripheral zone (PZ) DWI has been established has the dominant sequence for PCa [1]. DWI reflects and measures the diffusion of water molecules within biological tissues due to thermal Brownian motion. This Gaussian diffusion is quantified by an apparent diffusion coefficient (ADC) that depends on the fluctuation’s degree of the water molecules in interactions with cell compartments and macromolecules. A simple mono-exponential decay model characterizes this type of diffusion [3,4]. In the case of tumors, the diffusion is impeded due to increased cellularity of the neoplastic tissue leading to the hypointensity on the ADC map.
M. Valerio et al. / European Journal of Radiology 85 (2016) 790–794
It is now established that the ADC value reflects tumor aggressiveness because it is correlated by an inverse relationship to the Gleason score [1,5–7]. High b-values (≥1000 mm2 /s) have been demonstrated to increase the sensitivity for detection of high grade PCa, because reflecting the strength of the diffusion effects and to minimizing T2 shine-through and perfusion effects within the capillary networks [8]. However, the use of high b-values involves significant limitations such as reducing the image spatial resolution and decreasing SNR ratio. These limitations are partially related to the blood microcirculation network of the capillaries capable to change the intensity of the diffusion signal at very low values of b (<200 mm2 /s) [9]. This pseudo-diffusion based on microvascular perfusion was proposed to explain the theory of intravoxel incoherent motion (IVIM) [9,10]. In the IVIM model, the extravascular molecular diffusion and the microcirculation of blood within the capillaries can be separated using a bi-exponential decay function [11]. In the present study, we analyzed neoplastic – biopsy proved – and healthy tissues within the peripheral zone (PZ), using DWI sequence with multiple b-values and applying bi-exponential fits to the diffusion decay curves in order to calculate the molecular diffusion coefficient (D), perfusion-related diffusion coefficient (D*), and perfusion fraction (f). Primary aim of the study was to determine the possible impact of IVIM in the detection of PZ PCa. Secondary aim was to assess whether ADC and IVIM parameters are able to stratify the pathological grade of PZ PCa.
791
Average ADC (apparent diffusion coefficient) values were calculated by placing a region of interest (ROI) on the suspicious areas on the ADC maps obtained. IVIM parameters were acquired with 11 b-values (0, 10, 20, 30, 40, 50, 80, 100, 200, 400, 800 s/mm2 ). - Dynamic contrast-enhanced MRI obtained using a gradientecho T1-weighted sequence in the axial plane (TR, 3 ms; TE, 2 ms; thickness, 3 mm; time resolution, 12 sections/3 s; matrix, 320 × 192). 2.2. Image interpretation and data analysis MRI examinations were interpreted and analyzed by two radiologists in consensus, blinded to the patient’s clinical history and biopsy data. Index lesion was defined as the PZ area with higher probability of PCa according to PI-RADS v2 guidelines (Fig. 1) [1]. Regions of interest (ROIs) were placed in the index lesion areas, in the contralateral healthy PZ and in the clear benign prostate hypertrophy (BPH) zone in order to calculate the IVIM parameters in those sites. The ROIs were chosen to be as large as possible, consistent with minimal contamination from unintended tissues. The data analysis was performed with Olea Sphere software (Olea Medical, LaCiotat, France). To extract the two diffusion coefficients, one related to molecular diffusion restriction (D), another related to the tissue perfusion (D*) and finally the vascular volume fraction (f), the following biexponential equation was used: S(b) = S0 × [fe−bD∗ + (1 − f )−bD ][12]
2. Material and methods Our institutional review board approved this prospectic study and no informed consent was required. Patients with clinical suspect of PCa because serum PSA elevated (>2.5–4.0 ng/mL) and clinical positive digital-rectal exploration underwent to mp-MRI; fifty-three patients showed at least one lesion at the level of PZ with high probability of PCa according to PI-RADS v2 guidelines [1]. Multiparametric magnetic resonance (MR)—targeted biopsy (TB) with ultrasound fusion co-registered technique (Urostation, Koelis, France) was performed at the level of the PZ lesion with highest probability of PCa, at least 1 week after the mp-MR scan; a minimum of 2 up to 4 samples, depending on lesion size, were collected. Furthermore, a systematic biopsy with 12 specimens was performed on the same biopsy session. Biopsy results were considered the gold standard in the evaluation of prostate parenchyma.
2.1. MR acquisition protocols MR imaging of the pelvis was performed using a 3-T scan (Discovery MR750, GE Healthcare, Milwaukee, USA) equipped with an eight-channel torso phased-array coil and an endorectal coil. The MR imaging protocol of the prostate included the following sequences: - T2-weighted turbo spin echo sequences (repetition time (TR), 4500 ms; echo time (TE), 110 ms; thickness, 3 mm; matrix, 352 × 352) acquired in the axial, sagittal and coronal planes. - Diffusion-weighted (DW) sequences (slice thickness, 3 mm; TR, 3100 ms; TE, 102 ms; exponential b values of 0, 500, 1000 and 3000 s/mm2 ) acquired in the axial plane.
where S(b) is the signal intensity, S0 the signal reference and b is the b value. The lower and upper bounds for the parameters D, D* and f were 0–10 10−3 mm2 /s, 10–150 10−3 mm2 /s e 0–1, respectively. The apparent diffusion coefficient (ADC) was computed on the same ROIs by linear regression from the isotropic images from each b-value according to the mono-exponential equation: S(b) = S0 × e−bADC The lower and upper bounds for the parameters ADC was 0–100 10−3 mm2 /s. 2.3. Statistical analysis Inferential statistics analysis was performed in order to evaluate the single parameters of IVIM analisys. The results coming from mp-MR imaging measures were analyzed by means of classical inferential approaches. A paired t-test was performed to assess the statistical significance of the T2, ADC, D, D* and f parameters between cancerous and contralateral healthy tissues. On the other hand, a two-tailed independent samples t-test was used to check the statistical significance of the difference between low-grade (LG) and combined intermediate-high-grade (HG) tumors using mp-MR imaging measures. A p value of ≤0.05 was considered as the threshold of statistical significance for both paired and independent samples t-tests. Linear discriminant analysis was performed in order to evaluate the different diagnostic performace for PZ PCa diagnosis using three different datasets [13,14]: - the conventional parameters (T2 and ADC) - the IVIM parameters (D, D* and f) - combined T2-weighted imaging/DWI and IVIM parameters (T2, ADC, D, D* and f),
792
M. Valerio et al. / European Journal of Radiology 85 (2016) 790–794
Fig. 1. 49 year-old man with prostate cancer GS 7 (4 + 3) and serum PSA level of 8.2 ng/ml. (A) T2-Weighted anatomical image shows the tumor in posterior left PZ, as hypointense area (marked). (B) DWI b = 3000 s/mm2 ; (C) ADC map (b = 0, 500, 1000, 3000 s/mm2 ); (D) D map (coefficient of molecular diffusion), shows a well-circumscribed hypointense area corresponding to the tumor; (E) f map, the tumor represents an area of ill-defined hypointensity; (F) D* map (perfusion-related diffusion) the tumor appears as a well-circumscribed, recognizable, hyperintense area.
Table 1 T2 , apparent diffusion coefficient (ADC), pure molecular-based diffusion coefficient (D), perfusion-related diffusion (D*) and perfusion fraction (f) mean values (±SD) in ROIs of healthy peripheral zone and PCa. Healthy peripheral zone T2 (ms) ADC (×10−3 mm2 /s) D (×10−3 mm2 /s) D* (×10−3 mm2 /s) f (%)
3750.70 1.39 1.77 10.25 10.07
± ± ± ± ±
1735.37 0.48 0.36 10.52 9.70
Linear discriminant analysis (LDA) is a supervised pattern recognition method in which the analyzed variables pertain to two classes: the ‘diagnosis’ (dependent, Y) and ‘symptoms’ (independent, X) variables. The aim of LDA is to find the linear combination of X variables that explains the Y variable(s) better (i.e., to build an explicit criterion based on a linear function of Xs minimizing the errors of classification as for Y). This goal is achieved by the generation of a set of weights multiplying each X variable so to build a metrics in which the errors of assignment of each statistical unit to the correct Y class is minimized. SAS software (SAS Institute Inc., Cary, NC, USA) was used for all the statistical analysis.
3. Results The mean values of T2, ADC, D, D*, and f parameters measured in healthy PZ and PCa regions were 3750.70 ± 1735.37 vs. 2749.82 ± 1324.67 (ms, p < 0.00001), 1.39 ± 0.48 vs. 0.76 ± 0.27 (×10−3 mm2 /s, p < 0.00001), 1.77 ± 0.36 vs. 0.99 ± 0.38 (×10−3 mm2 /s, p < 0.00001), 10.25 ± 10.52 vs. 15.56 ± 12.91 (×10−3 mm2 /s, p = 0.01), and 10.07 ± 9.70 vs. 9.35 ± 5.97 (%, p = 0.26), respectively (Table 1).
PCa 2749.82 0.76 0.99 15.56 9.35
± ± ± ± ±
1324.67 0.27 0.38 12.91 5.97
p value
t value
<0.00001 <0.00001 <0.00001 0.01 0.26
−2.94 −6.43 −8.95 1.51 −0.17
We found that the values of T2, ADC, and D parameters were significantly reduced in the index lesion area compared to the ones of healthy tissues. The D* parameter was significantly increased in the index lesion area compare to the healthy PZ; no significant difference between f measurements in the healthy PZ and index lesion areas was found. Table 1 shows, the t values: D, ADC and T2 (in this order) are the parameters with the greater discriminatory power compared to the others. Furthermore, since the combination of two or more parameters has an overall efficiency not equal to the sum of the discriminatory powers of each parameter, we subsequently decided to perform linear discriminant analysis (LDA) on different parameters combinations. In particular, the discriminant function was computed for the conventional parameters (T2 and ADC), the IVIM parameters (D, D* and f) and the combined T2-weighted imaging/DWI and IVIM parameters (T2, ADC, D, D* and f), in order to evaluate if the additional use of IVIM imaging could improve the diagnosis of PCa. The specificity, sensitivity and accuracy of T2-weighted imaging together with DWI and IVIM parameters were 100, 96 and 98%, respectively, in comparison with 88, 92 and 90% and 96, 92 and 94 for T2-weighted imaging/DWI and IVIM alone (Table 2). Therefore, the additional use of IVIM increased the specificity, sensitivity and
M. Valerio et al. / European Journal of Radiology 85 (2016) 790–794 Table 2 Discriminant function analysis comparing healthy PZ vs. PCa with regard to ADC and T2 or D*, D and f, or all parameters. For each discriminant function the specificity, sensitivity and accuracy are reported in percent units.
F-value Specificity Sensitivity Accuracy
T2 and ADC
D, D* and f
15.23 88 92 90
15.60 96 92 94
T2 , ADC, D, D* and f 12.04 100 96 98
Table 3 Correlation of tumour Gleason grade (low grade (≤6) and high grade (>6)) to DWI and IVIM parameters (t-test). Variable
LG (Mean ± SD)
T2 ADC D D* f
2774.53 0.87 1.17 11.07 9.6
± ± ± ± ±
1068.0 0.28 0.33 9.07 5.6
HG (Mean ± SD) 2716.13 0.61 0.75 21.68 8.9
± ± ± ± ±
1632.98 0.12 0.29 14.94 6.5
p value 0.877 0.0001 <0.00001 0.003 0.690
accuracy of conventional T2-weighted imaging/DWI for predicting prostate cancer detection (Table 2). Subsequently, we used a t-test to explore the possibility of T2, DWI and IVIM parameters to distinguish low-grade (LG, Gleason score ≤6) from intermediate/high-grade (HG, Gleason score >6) PCa. In our population we experienced 30 patients with LG PCa (Gleason score = 6) and 23 patients with HG PCa (27 patients with Gleason score = 7 and 6 patients with Gleason score = 8). ADC (p = 0.0001), D (p < 0.00001) and D* (p = 0.003) values were correlated with tumor Gleason scores (Table 3). 4. Discussion DWI is currently considered a key component of prostate mpMRI examinations, especially in the PZ where it has been established as dominant sequence for the detection of PCa(0). However, to improve the significance of DWI, it is necessary to evaluate separately the two components of diffusion: the pure molecular diffusion and the perfusion-related diffusion (pseudodiffusion) originating from capillary microcirculation. IVIM-DWI, applying a bi-exponential fitting function, allows the extraction of pure molecular diffusion parameters (D) and perfusion-related diffusion parameters (D* and f). Initially, the application of IVIM imaging has shown potentials in the evaluation of pancreatic lesions, ureteral obstruction and liver cirrhosis [15–17]. Recently, few authors reported interesting results of IVIM model in the study of prostate cancer [18–21]. In our study we first compared the T2-weighted imaging/DWI with IVIM parameters (T2, ADC, D, D*, f) obtained from the index lesion and healthy PZ areas in order to evaluate if there are any significant differences (Table 1). The T2, ADC and D values were significantly lower in the index lesion areas compared to those of the healthy PZ areas (p < 0.00001). Similar results have been already published literature suggesting that cellular structural changes occur in premalignant and malignant lesions [18–21]. In the present study, we also found a statistical significant difference of D* parameter calculated in the index area compared with healthy PZ (p = 0.01). This result is completely new compared with previous studies [15,16]; however, we observed a large standard deviation (SDs) in the evaluation of D* parameter. Furthermore, in our study, there was no significant difference between f measurements in the index area compared with healthy PZ (p = 0.26); we observed a highly variability of the f parameter for healthy PZ areas compare to the index lesions areas. In line with our
793
results, also Luciani et al. reported there was no significant difference between f measurements in the neoplastic and healthy tissue [17]; however, several studies showed different results for f parameter: both higher and lower values were reported [18,19,21–25]. Secondly, in order to evaluate the different diagnostic performace for PZ PCa diagnosis we performed the discriminant analysis on the conventional parameters (T2 and ADC), the IVIM parameters (D, D* and f) and the combined T2-weighted imaging/DWI and IVIM parameters (T2, ADC, D, D* and f), The results showed that IVIM parameters combined to conventional protocol increased the specificity from 88% to 100%, the sensitivity from 92% to 96% and the accuracy from 90% to 98% compared to the conventional protocol (Table 2). The results were correlated with biopsy in order to evaluate if there is a possible relation between Gleason score and IVIM and conventional parameters. The results demonstrated there is a statistical difference between low-grade tumors (≤6 GS) and intermediate/high-grade tumors (>6 GS) in that ADC, D and D* dataset (p = 0.001, p < 0.0001 and p = 0.003 respectively). In particular, D has been demonstrated to have the highest significativity. Also Zhang et al. analyzed the possibility to stratify the pathological grade of PCa by IVIM with histogram metrics; the study highlighted that ADC and D were significantly decreased in highgrade tumors, while D* and f did not have statistically significant modifications [25]. Our study has several limitations. First, we used mp-MR TB and the systematic biopsy to assess the PCa, whereas in other study prostatectomy results were considered as gold standard; this approach could allow a more precise differentiation between the PCa and healthy PZ groups. Second, our findings may be influenced by selection biases due to a small sample size and number of HG tumors especially with Gleason score ≥8. In order to overcome these limitations, an extended dataset and a discrimination procedure with the examination of a much broader (and heterogeneous) set of patients could be evocated; a rigorous (e.g., blind clinical study) study would be mandatory in order to standardize the criteria. 5. Conclusions In conclusion, our results suggest that adding DW-IVIM to the standard mp-MRI acquisition protocol could increase the specificity as well as the sensitivity, reaching higher diagnostic accuracy. For what that concerns, a statistical difference between LG tumors and HG has been demonstrated in that ADC, D and D* dataset; in particular, D has been found to have the highest significativity leading to possibility to use it as an index to distinguish low from intermediate/high-grade tumors. Conflict of interest We wish to confirm that there are no known conflicts of interest associated with this publication and there has been no significant financial support for this work that could have influenced its outcome. References [1] J.C. Weinreb, J.O. Barentsz, P.L. Choyke, et al. PI-RADS Prostate Imaging − Reporting and Data System: 2015, Version 2, Eur Urol. (September) (2015). [2] F. Panebianco, V. Barchetti, A. Sciarra, et al., Multiparametric magnetic resonance imaging vs. standard care in men being evaluated for prostate cancer: a randomized study, Urol. Oncol. 33 (2015) 1–7. [3] A.R. Padhani, G. Liu, D.M. Koh, et al., Diffusion weighted magnetic resonance imaging as a cancer biomarker: consensus and recommendations, Neoplasia 11 (2009) 102–125.
794
M. Valerio et al. / European Journal of Radiology 85 (2016) 790–794
[4] J.H. Jensen, J.A. Helpern, MRI quantification of non-Gaussian water diffusion by kurtosis analysis, NMR Biomed. 23 (2010) 698–710. [5] T. Hambrock, D.M. Somford, H.J. Huisman, et al., Relationship between apparent diffusion coefficients at 3.0-T MR imaging and Gleason grade in peripheral zone prostate cancer, Radiology 259 (2011) 453–461. [6] H.A. Vargas, O. Akin, T. Franiel, et al., Diffusion-weighted endorectal MR imaging at 3 T for prostate cancer: tumor detection and assessment of aggressiveness, Radiology 259 (2011) 775–784. [7] B. Turkbey, V.P. Shah, Y. Pang, et al., Is apparent diffusion coefficient associated with clinical risk scores for prostate cancers that are visible on 3-T MR images, Radiology 258 (2011) 488–495. [8] A.B. Rosenkrantz, N. Hindman, R.P. Lim, et al., Diffusion-weighted imaging of the prostate: comparison of b1000 and b2000 image sets for index lesion detection, J. Magn. Reson. Imaging 38 (2013) 694–700. [9] D. Le Bihan, E. Breton, D. Lallemand, et al., Separation of diffusion and perfusion in intravoxel incoherent motion MR imaging, Radiology 68 (1988) 497–505. [10] D. Le Bihan, Intravoxel incoherent motion perfusion MR imaging: a wake-up call, Radiology 249 (2008) 748–752. [11] R.V. Mulkern, A.S. Barnes, S.J. Haker, et al., Biexponential characterization of prostate tissue water diffusion decay curves over an extended b-factor range, Magn. Reson. Imaging 24 (2006) 563–568. [12] D. Le Bihan, E. Breton, D. Lallemand, et al., Separation of diffusion and perfusion in intravoxel incoherent motion MR imaging, Radiology 168 (2) (1988) 497–505. [13] R.A. Fisher, The use of multiple measurements in taxonomic problems, Ann. Eugen. 7 (2) (1936) 179–188. [14] T.W. Anderson, An Introduction to Multivariate Statistical Analysis, 2nd ed., John Wiley and Sons, New York, 1984. [15] A. Lemke, F.B. Laun, M. Klau, et al., Differentiation of pancreas carcinoma from healthy pancreatic tissue using multiple b-values: comparison of apparent diffusion coefficient and intravoxel incoherent motion derived parameters, Invest. Radiol. 44 (2009) 769–775.
[16] H.C. Thoeny, T. Binser, B. Roth, et al., Noninvasive assessment of acute ureteral obstruction with diffusion-weighted MR imaging: a prospective study, Radiology 252 (2009) 721–728. [17] A. Luciani, A. Vignaud, M. Cavet, et al., Liver cirrhosis: intravoxel incoherent motion MR imaging-pilot study, Radiology 249 (2008) 891–899. [18] J. Döpfert, A. Lemke, A. Weidner, L.R. Schad, Investigation of prostate cancer using diffusion-weighted intravoxel incoherent motion imaging, Magn. Reson. Imaging 29 (2011) 1053–1058. [19] H. Shinmoto, C. Tamura, S. Soga, et al., An intravoxel incoherent motion diffusion-weighted imaging study of prostate cancer, Am. J. Roentgenol. 4 (2012) 496–500. [20] S.F. Riches, K. Hawtin, E.M. Charles-Edwards, N.M. de Souza, Diffusion-weighted imaging of the prostate and rectal wall: comparison of biexponential and monoexponential modelled diffusion and associated perfusion coefficients, NMR Biomed. 22 (2009) 318–325. [21] J. Patel, E.E. Sigmund, H. Rusinek, Babb J.S. OeiM, B. Taouli, Diagnosis of cirrhosis with intravoxel incoherent motion diffusion MRI and dynamic contrast-enhanced MRI alone and in combination: preliminary experience, J. Magn. Reson. Imaging 31 (2010) 589–600. [22] Y. Pang, B. Turkbey, M. Bernardo, et al., Intravoxel incoherent motion MR imaging for prostate cancer: an evaluation of perfusion fraction and diffusion coefficient derived from different b-value combinations, Magn. Reson. Med. 69 (2013) 553–562. [23] I. Ocak, M. Bernardo, G. Metzger, et al., Dynamic contrast-enhanced MRI of prostate cancer at 3 T: a study of pharmacokinetic parameters, Am. J. Roentgenol. 189 (2007) 849–859. [24] Y. Mazaheri, H.A. Vargas, O. Akin, D.A. Goldman, H. Hricak, Reducing the influence of b-value selection on diffusion-weighted imaging of the prostate: evaluation of a revised monoexponential model within a clinical setting, J. Magn. Reson. Imaging 35 (2012) 660–668. [25] Y.D. Zhang, Q. Wang, C.J. Wu, et al., The histogram analysis of diffusion-weighted intravoxel incoherent motion (IVIM) imaging for differentiating the gleason grade of prostate cancer, Eur. Radiol. 25 (2015) 994–1004.
Contents lists available at ScienceDirect
European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad
3T multiparametric MRI of the prostate: Does intravoxel incoherent motion diffusion imaging have a role in the detection and stratification of prostate cancer in the peripheral zone? Mariacristina Valerio a , Chiara Zini a , Davide Fierro a , Francesca Giura a , Anna Colarieti a , Alessandro Giuliani b , Andrea Laghi a , Carlo Catalano a , Valeria Panebianco a,∗ a b
Dept. of Radiological Sciences, Oncology & Pathology—Sapienza University of Rome, V.le Regina Elena, 324, 00161 Roma, Italy Environment and Health Dept., Istituto Superiore di Sanità, Rome, V.le Regina Elena, 299, 00161 Roma, Italy
a r t i c l e
i n f o
Article history: Received 30 October 2015 Received in revised form 13 January 2016 Accepted 16 January 2016 Keywords: Multiparametric magnetic resonance imaging Diffusion weighted imaging Intravoxel incoherent motion Prostate cancer Peripheral zone Gleason score
a b s t r a c t Purpose: To evaluate the potential added value of the intravoxel incoherent motion model to conventional multiparametric magnetic resonance protocol in order to differentiate between healthy and neoplastic prostate tissue in the peripheral zone. Material and methods: Mono-exponential and bi-exponential fits were used to calculate ADC and IVIM parameters in 53 patients with peripheral zone biopsy proved tumor. Inferential statistics analysis was performed on T2, ADC and IVIM parameters (D, D*, f) comparing healthy and neoplastic tissues. Linear discriminant analysis was performed for the conventional parameters (T2 and ADC), the IVIM parameters (molecular diffusion coefficient (D), perfusion-related diffusion coefficient (D*), and perfusion fraction (f) and the combined T2-weighted imaging/DWI and IVIM parameters (T2, ADC, D, D* and f). A correlation with Gleason scores was achieved. Results: The values of T2, ADC and D were significantly lower in cancerous tissues (2749.82 ± 1324.67 ms, 0.76 ± 0.27 × 10−3 mm2 /s and 0.99 ± 0.38 × 10−3 mm2 /s respectively) compared to those found in the healthy tissues (3750.70 ± 1735.37 ms, 1.39 ± 0.48 × 10−3 mm2 /s and 1.77 ± 0.36 × 10−3 mm2 /s respectively); D* parameter was significantly increased in neoplastic compared to healthy tissue (15.56 ± 12.91 × 10−3 mm2 /s and 10.25 ± 10.52 × 10−3 mm2 /s respectively). The specificity, sensitivity and accuracy of the T2-weighted imaging/DWI and IVIM parameters were 100, 96 and 98%, respectively, compare to 88, 92 and 90% and 96, 92 and 94 for T2-weighted imaging/ADC and IVIM alone. Conclusions: IVIM parameters increase the specificity and sensitivity in the evaluation of peripheral zone prostate cancer. A statistical difference between low grade tumors and high grade tumors has been demostrated in that ADC, D and D* dataset; in particular, D has been found to have the highest significativity. © 2016 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Multiparametric magnetic resonance (mp-MR) combined with magnetic resonance (MR)—targeted biopsy is considered valuable tool for early detection of clinically significant prostate cancer (PCa) [1]. In a recent randomized study comparing mp-MRI/biopsy vs. TRUS-guided biopsy, accuracy of mp-MRI in detecting clinically significant PCa was 97% [2].
∗ Corresponding author. E-mail addresses: [email protected], [email protected] (V. Panebianco). http://dx.doi.org/10.1016/j.ejrad.2016.01.006 0720-048X/© 2016 Elsevier Ireland Ltd. All rights reserved.
Among the various sequences composing the mp-MR acquisition protocol, diffusion-weighted imaging (DWI) has been claimed to contribute to T2-weighted MRI in the detection and localization of PCa; in particular, in the peripheral zone (PZ) DWI has been established has the dominant sequence for PCa [1]. DWI reflects and measures the diffusion of water molecules within biological tissues due to thermal Brownian motion. This Gaussian diffusion is quantified by an apparent diffusion coefficient (ADC) that depends on the fluctuation’s degree of the water molecules in interactions with cell compartments and macromolecules. A simple mono-exponential decay model characterizes this type of diffusion [3,4]. In the case of tumors, the diffusion is impeded due to increased cellularity of the neoplastic tissue leading to the hypointensity on the ADC map.
M. Valerio et al. / European Journal of Radiology 85 (2016) 790–794
It is now established that the ADC value reflects tumor aggressiveness because it is correlated by an inverse relationship to the Gleason score [1,5–7]. High b-values (≥1000 mm2 /s) have been demonstrated to increase the sensitivity for detection of high grade PCa, because reflecting the strength of the diffusion effects and to minimizing T2 shine-through and perfusion effects within the capillary networks [8]. However, the use of high b-values involves significant limitations such as reducing the image spatial resolution and decreasing SNR ratio. These limitations are partially related to the blood microcirculation network of the capillaries capable to change the intensity of the diffusion signal at very low values of b (<200 mm2 /s) [9]. This pseudo-diffusion based on microvascular perfusion was proposed to explain the theory of intravoxel incoherent motion (IVIM) [9,10]. In the IVIM model, the extravascular molecular diffusion and the microcirculation of blood within the capillaries can be separated using a bi-exponential decay function [11]. In the present study, we analyzed neoplastic – biopsy proved – and healthy tissues within the peripheral zone (PZ), using DWI sequence with multiple b-values and applying bi-exponential fits to the diffusion decay curves in order to calculate the molecular diffusion coefficient (D), perfusion-related diffusion coefficient (D*), and perfusion fraction (f). Primary aim of the study was to determine the possible impact of IVIM in the detection of PZ PCa. Secondary aim was to assess whether ADC and IVIM parameters are able to stratify the pathological grade of PZ PCa.
791
Average ADC (apparent diffusion coefficient) values were calculated by placing a region of interest (ROI) on the suspicious areas on the ADC maps obtained. IVIM parameters were acquired with 11 b-values (0, 10, 20, 30, 40, 50, 80, 100, 200, 400, 800 s/mm2 ). - Dynamic contrast-enhanced MRI obtained using a gradientecho T1-weighted sequence in the axial plane (TR, 3 ms; TE, 2 ms; thickness, 3 mm; time resolution, 12 sections/3 s; matrix, 320 × 192). 2.2. Image interpretation and data analysis MRI examinations were interpreted and analyzed by two radiologists in consensus, blinded to the patient’s clinical history and biopsy data. Index lesion was defined as the PZ area with higher probability of PCa according to PI-RADS v2 guidelines (Fig. 1) [1]. Regions of interest (ROIs) were placed in the index lesion areas, in the contralateral healthy PZ and in the clear benign prostate hypertrophy (BPH) zone in order to calculate the IVIM parameters in those sites. The ROIs were chosen to be as large as possible, consistent with minimal contamination from unintended tissues. The data analysis was performed with Olea Sphere software (Olea Medical, LaCiotat, France). To extract the two diffusion coefficients, one related to molecular diffusion restriction (D), another related to the tissue perfusion (D*) and finally the vascular volume fraction (f), the following biexponential equation was used: S(b) = S0 × [fe−bD∗ + (1 − f )−bD ][12]
2. Material and methods Our institutional review board approved this prospectic study and no informed consent was required. Patients with clinical suspect of PCa because serum PSA elevated (>2.5–4.0 ng/mL) and clinical positive digital-rectal exploration underwent to mp-MRI; fifty-three patients showed at least one lesion at the level of PZ with high probability of PCa according to PI-RADS v2 guidelines [1]. Multiparametric magnetic resonance (MR)—targeted biopsy (TB) with ultrasound fusion co-registered technique (Urostation, Koelis, France) was performed at the level of the PZ lesion with highest probability of PCa, at least 1 week after the mp-MR scan; a minimum of 2 up to 4 samples, depending on lesion size, were collected. Furthermore, a systematic biopsy with 12 specimens was performed on the same biopsy session. Biopsy results were considered the gold standard in the evaluation of prostate parenchyma.
2.1. MR acquisition protocols MR imaging of the pelvis was performed using a 3-T scan (Discovery MR750, GE Healthcare, Milwaukee, USA) equipped with an eight-channel torso phased-array coil and an endorectal coil. The MR imaging protocol of the prostate included the following sequences: - T2-weighted turbo spin echo sequences (repetition time (TR), 4500 ms; echo time (TE), 110 ms; thickness, 3 mm; matrix, 352 × 352) acquired in the axial, sagittal and coronal planes. - Diffusion-weighted (DW) sequences (slice thickness, 3 mm; TR, 3100 ms; TE, 102 ms; exponential b values of 0, 500, 1000 and 3000 s/mm2 ) acquired in the axial plane.
where S(b) is the signal intensity, S0 the signal reference and b is the b value. The lower and upper bounds for the parameters D, D* and f were 0–10 10−3 mm2 /s, 10–150 10−3 mm2 /s e 0–1, respectively. The apparent diffusion coefficient (ADC) was computed on the same ROIs by linear regression from the isotropic images from each b-value according to the mono-exponential equation: S(b) = S0 × e−bADC The lower and upper bounds for the parameters ADC was 0–100 10−3 mm2 /s. 2.3. Statistical analysis Inferential statistics analysis was performed in order to evaluate the single parameters of IVIM analisys. The results coming from mp-MR imaging measures were analyzed by means of classical inferential approaches. A paired t-test was performed to assess the statistical significance of the T2, ADC, D, D* and f parameters between cancerous and contralateral healthy tissues. On the other hand, a two-tailed independent samples t-test was used to check the statistical significance of the difference between low-grade (LG) and combined intermediate-high-grade (HG) tumors using mp-MR imaging measures. A p value of ≤0.05 was considered as the threshold of statistical significance for both paired and independent samples t-tests. Linear discriminant analysis was performed in order to evaluate the different diagnostic performace for PZ PCa diagnosis using three different datasets [13,14]: - the conventional parameters (T2 and ADC) - the IVIM parameters (D, D* and f) - combined T2-weighted imaging/DWI and IVIM parameters (T2, ADC, D, D* and f),
792
M. Valerio et al. / European Journal of Radiology 85 (2016) 790–794
Fig. 1. 49 year-old man with prostate cancer GS 7 (4 + 3) and serum PSA level of 8.2 ng/ml. (A) T2-Weighted anatomical image shows the tumor in posterior left PZ, as hypointense area (marked). (B) DWI b = 3000 s/mm2 ; (C) ADC map (b = 0, 500, 1000, 3000 s/mm2 ); (D) D map (coefficient of molecular diffusion), shows a well-circumscribed hypointense area corresponding to the tumor; (E) f map, the tumor represents an area of ill-defined hypointensity; (F) D* map (perfusion-related diffusion) the tumor appears as a well-circumscribed, recognizable, hyperintense area.
Table 1 T2 , apparent diffusion coefficient (ADC), pure molecular-based diffusion coefficient (D), perfusion-related diffusion (D*) and perfusion fraction (f) mean values (±SD) in ROIs of healthy peripheral zone and PCa. Healthy peripheral zone T2 (ms) ADC (×10−3 mm2 /s) D (×10−3 mm2 /s) D* (×10−3 mm2 /s) f (%)
3750.70 1.39 1.77 10.25 10.07
± ± ± ± ±
1735.37 0.48 0.36 10.52 9.70
Linear discriminant analysis (LDA) is a supervised pattern recognition method in which the analyzed variables pertain to two classes: the ‘diagnosis’ (dependent, Y) and ‘symptoms’ (independent, X) variables. The aim of LDA is to find the linear combination of X variables that explains the Y variable(s) better (i.e., to build an explicit criterion based on a linear function of Xs minimizing the errors of classification as for Y). This goal is achieved by the generation of a set of weights multiplying each X variable so to build a metrics in which the errors of assignment of each statistical unit to the correct Y class is minimized. SAS software (SAS Institute Inc., Cary, NC, USA) was used for all the statistical analysis.
3. Results The mean values of T2, ADC, D, D*, and f parameters measured in healthy PZ and PCa regions were 3750.70 ± 1735.37 vs. 2749.82 ± 1324.67 (ms, p < 0.00001), 1.39 ± 0.48 vs. 0.76 ± 0.27 (×10−3 mm2 /s, p < 0.00001), 1.77 ± 0.36 vs. 0.99 ± 0.38 (×10−3 mm2 /s, p < 0.00001), 10.25 ± 10.52 vs. 15.56 ± 12.91 (×10−3 mm2 /s, p = 0.01), and 10.07 ± 9.70 vs. 9.35 ± 5.97 (%, p = 0.26), respectively (Table 1).
PCa 2749.82 0.76 0.99 15.56 9.35
± ± ± ± ±
1324.67 0.27 0.38 12.91 5.97
p value
t value
<0.00001 <0.00001 <0.00001 0.01 0.26
−2.94 −6.43 −8.95 1.51 −0.17
We found that the values of T2, ADC, and D parameters were significantly reduced in the index lesion area compared to the ones of healthy tissues. The D* parameter was significantly increased in the index lesion area compare to the healthy PZ; no significant difference between f measurements in the healthy PZ and index lesion areas was found. Table 1 shows, the t values: D, ADC and T2 (in this order) are the parameters with the greater discriminatory power compared to the others. Furthermore, since the combination of two or more parameters has an overall efficiency not equal to the sum of the discriminatory powers of each parameter, we subsequently decided to perform linear discriminant analysis (LDA) on different parameters combinations. In particular, the discriminant function was computed for the conventional parameters (T2 and ADC), the IVIM parameters (D, D* and f) and the combined T2-weighted imaging/DWI and IVIM parameters (T2, ADC, D, D* and f), in order to evaluate if the additional use of IVIM imaging could improve the diagnosis of PCa. The specificity, sensitivity and accuracy of T2-weighted imaging together with DWI and IVIM parameters were 100, 96 and 98%, respectively, in comparison with 88, 92 and 90% and 96, 92 and 94 for T2-weighted imaging/DWI and IVIM alone (Table 2). Therefore, the additional use of IVIM increased the specificity, sensitivity and
M. Valerio et al. / European Journal of Radiology 85 (2016) 790–794 Table 2 Discriminant function analysis comparing healthy PZ vs. PCa with regard to ADC and T2 or D*, D and f, or all parameters. For each discriminant function the specificity, sensitivity and accuracy are reported in percent units.
F-value Specificity Sensitivity Accuracy
T2 and ADC
D, D* and f
15.23 88 92 90
15.60 96 92 94
T2 , ADC, D, D* and f 12.04 100 96 98
Table 3 Correlation of tumour Gleason grade (low grade (≤6) and high grade (>6)) to DWI and IVIM parameters (t-test). Variable
LG (Mean ± SD)
T2 ADC D D* f
2774.53 0.87 1.17 11.07 9.6
± ± ± ± ±
1068.0 0.28 0.33 9.07 5.6
HG (Mean ± SD) 2716.13 0.61 0.75 21.68 8.9
± ± ± ± ±
1632.98 0.12 0.29 14.94 6.5
p value 0.877 0.0001 <0.00001 0.003 0.690
accuracy of conventional T2-weighted imaging/DWI for predicting prostate cancer detection (Table 2). Subsequently, we used a t-test to explore the possibility of T2, DWI and IVIM parameters to distinguish low-grade (LG, Gleason score ≤6) from intermediate/high-grade (HG, Gleason score >6) PCa. In our population we experienced 30 patients with LG PCa (Gleason score = 6) and 23 patients with HG PCa (27 patients with Gleason score = 7 and 6 patients with Gleason score = 8). ADC (p = 0.0001), D (p < 0.00001) and D* (p = 0.003) values were correlated with tumor Gleason scores (Table 3). 4. Discussion DWI is currently considered a key component of prostate mpMRI examinations, especially in the PZ where it has been established as dominant sequence for the detection of PCa(0). However, to improve the significance of DWI, it is necessary to evaluate separately the two components of diffusion: the pure molecular diffusion and the perfusion-related diffusion (pseudodiffusion) originating from capillary microcirculation. IVIM-DWI, applying a bi-exponential fitting function, allows the extraction of pure molecular diffusion parameters (D) and perfusion-related diffusion parameters (D* and f). Initially, the application of IVIM imaging has shown potentials in the evaluation of pancreatic lesions, ureteral obstruction and liver cirrhosis [15–17]. Recently, few authors reported interesting results of IVIM model in the study of prostate cancer [18–21]. In our study we first compared the T2-weighted imaging/DWI with IVIM parameters (T2, ADC, D, D*, f) obtained from the index lesion and healthy PZ areas in order to evaluate if there are any significant differences (Table 1). The T2, ADC and D values were significantly lower in the index lesion areas compared to those of the healthy PZ areas (p < 0.00001). Similar results have been already published literature suggesting that cellular structural changes occur in premalignant and malignant lesions [18–21]. In the present study, we also found a statistical significant difference of D* parameter calculated in the index area compared with healthy PZ (p = 0.01). This result is completely new compared with previous studies [15,16]; however, we observed a large standard deviation (SDs) in the evaluation of D* parameter. Furthermore, in our study, there was no significant difference between f measurements in the index area compared with healthy PZ (p = 0.26); we observed a highly variability of the f parameter for healthy PZ areas compare to the index lesions areas. In line with our
793
results, also Luciani et al. reported there was no significant difference between f measurements in the neoplastic and healthy tissue [17]; however, several studies showed different results for f parameter: both higher and lower values were reported [18,19,21–25]. Secondly, in order to evaluate the different diagnostic performace for PZ PCa diagnosis we performed the discriminant analysis on the conventional parameters (T2 and ADC), the IVIM parameters (D, D* and f) and the combined T2-weighted imaging/DWI and IVIM parameters (T2, ADC, D, D* and f), The results showed that IVIM parameters combined to conventional protocol increased the specificity from 88% to 100%, the sensitivity from 92% to 96% and the accuracy from 90% to 98% compared to the conventional protocol (Table 2). The results were correlated with biopsy in order to evaluate if there is a possible relation between Gleason score and IVIM and conventional parameters. The results demonstrated there is a statistical difference between low-grade tumors (≤6 GS) and intermediate/high-grade tumors (>6 GS) in that ADC, D and D* dataset (p = 0.001, p < 0.0001 and p = 0.003 respectively). In particular, D has been demonstrated to have the highest significativity. Also Zhang et al. analyzed the possibility to stratify the pathological grade of PCa by IVIM with histogram metrics; the study highlighted that ADC and D were significantly decreased in highgrade tumors, while D* and f did not have statistically significant modifications [25]. Our study has several limitations. First, we used mp-MR TB and the systematic biopsy to assess the PCa, whereas in other study prostatectomy results were considered as gold standard; this approach could allow a more precise differentiation between the PCa and healthy PZ groups. Second, our findings may be influenced by selection biases due to a small sample size and number of HG tumors especially with Gleason score ≥8. In order to overcome these limitations, an extended dataset and a discrimination procedure with the examination of a much broader (and heterogeneous) set of patients could be evocated; a rigorous (e.g., blind clinical study) study would be mandatory in order to standardize the criteria. 5. Conclusions In conclusion, our results suggest that adding DW-IVIM to the standard mp-MRI acquisition protocol could increase the specificity as well as the sensitivity, reaching higher diagnostic accuracy. For what that concerns, a statistical difference between LG tumors and HG has been demonstrated in that ADC, D and D* dataset; in particular, D has been found to have the highest significativity leading to possibility to use it as an index to distinguish low from intermediate/high-grade tumors. Conflict of interest We wish to confirm that there are no known conflicts of interest associated with this publication and there has been no significant financial support for this work that could have influenced its outcome. References [1] J.C. Weinreb, J.O. Barentsz, P.L. Choyke, et al. PI-RADS Prostate Imaging − Reporting and Data System: 2015, Version 2, Eur Urol. (September) (2015). [2] F. Panebianco, V. Barchetti, A. Sciarra, et al., Multiparametric magnetic resonance imaging vs. standard care in men being evaluated for prostate cancer: a randomized study, Urol. Oncol. 33 (2015) 1–7. [3] A.R. Padhani, G. Liu, D.M. Koh, et al., Diffusion weighted magnetic resonance imaging as a cancer biomarker: consensus and recommendations, Neoplasia 11 (2009) 102–125.
794
M. Valerio et al. / European Journal of Radiology 85 (2016) 790–794
[4] J.H. Jensen, J.A. Helpern, MRI quantification of non-Gaussian water diffusion by kurtosis analysis, NMR Biomed. 23 (2010) 698–710. [5] T. Hambrock, D.M. Somford, H.J. Huisman, et al., Relationship between apparent diffusion coefficients at 3.0-T MR imaging and Gleason grade in peripheral zone prostate cancer, Radiology 259 (2011) 453–461. [6] H.A. Vargas, O. Akin, T. Franiel, et al., Diffusion-weighted endorectal MR imaging at 3 T for prostate cancer: tumor detection and assessment of aggressiveness, Radiology 259 (2011) 775–784. [7] B. Turkbey, V.P. Shah, Y. Pang, et al., Is apparent diffusion coefficient associated with clinical risk scores for prostate cancers that are visible on 3-T MR images, Radiology 258 (2011) 488–495. [8] A.B. Rosenkrantz, N. Hindman, R.P. Lim, et al., Diffusion-weighted imaging of the prostate: comparison of b1000 and b2000 image sets for index lesion detection, J. Magn. Reson. Imaging 38 (2013) 694–700. [9] D. Le Bihan, E. Breton, D. Lallemand, et al., Separation of diffusion and perfusion in intravoxel incoherent motion MR imaging, Radiology 68 (1988) 497–505. [10] D. Le Bihan, Intravoxel incoherent motion perfusion MR imaging: a wake-up call, Radiology 249 (2008) 748–752. [11] R.V. Mulkern, A.S. Barnes, S.J. Haker, et al., Biexponential characterization of prostate tissue water diffusion decay curves over an extended b-factor range, Magn. Reson. Imaging 24 (2006) 563–568. [12] D. Le Bihan, E. Breton, D. Lallemand, et al., Separation of diffusion and perfusion in intravoxel incoherent motion MR imaging, Radiology 168 (2) (1988) 497–505. [13] R.A. Fisher, The use of multiple measurements in taxonomic problems, Ann. Eugen. 7 (2) (1936) 179–188. [14] T.W. Anderson, An Introduction to Multivariate Statistical Analysis, 2nd ed., John Wiley and Sons, New York, 1984. [15] A. Lemke, F.B. Laun, M. Klau, et al., Differentiation of pancreas carcinoma from healthy pancreatic tissue using multiple b-values: comparison of apparent diffusion coefficient and intravoxel incoherent motion derived parameters, Invest. Radiol. 44 (2009) 769–775.
[16] H.C. Thoeny, T. Binser, B. Roth, et al., Noninvasive assessment of acute ureteral obstruction with diffusion-weighted MR imaging: a prospective study, Radiology 252 (2009) 721–728. [17] A. Luciani, A. Vignaud, M. Cavet, et al., Liver cirrhosis: intravoxel incoherent motion MR imaging-pilot study, Radiology 249 (2008) 891–899. [18] J. Döpfert, A. Lemke, A. Weidner, L.R. Schad, Investigation of prostate cancer using diffusion-weighted intravoxel incoherent motion imaging, Magn. Reson. Imaging 29 (2011) 1053–1058. [19] H. Shinmoto, C. Tamura, S. Soga, et al., An intravoxel incoherent motion diffusion-weighted imaging study of prostate cancer, Am. J. Roentgenol. 4 (2012) 496–500. [20] S.F. Riches, K. Hawtin, E.M. Charles-Edwards, N.M. de Souza, Diffusion-weighted imaging of the prostate and rectal wall: comparison of biexponential and monoexponential modelled diffusion and associated perfusion coefficients, NMR Biomed. 22 (2009) 318–325. [21] J. Patel, E.E. Sigmund, H. Rusinek, Babb J.S. OeiM, B. Taouli, Diagnosis of cirrhosis with intravoxel incoherent motion diffusion MRI and dynamic contrast-enhanced MRI alone and in combination: preliminary experience, J. Magn. Reson. Imaging 31 (2010) 589–600. [22] Y. Pang, B. Turkbey, M. Bernardo, et al., Intravoxel incoherent motion MR imaging for prostate cancer: an evaluation of perfusion fraction and diffusion coefficient derived from different b-value combinations, Magn. Reson. Med. 69 (2013) 553–562. [23] I. Ocak, M. Bernardo, G. Metzger, et al., Dynamic contrast-enhanced MRI of prostate cancer at 3 T: a study of pharmacokinetic parameters, Am. J. Roentgenol. 189 (2007) 849–859. [24] Y. Mazaheri, H.A. Vargas, O. Akin, D.A. Goldman, H. Hricak, Reducing the influence of b-value selection on diffusion-weighted imaging of the prostate: evaluation of a revised monoexponential model within a clinical setting, J. Magn. Reson. Imaging 35 (2012) 660–668. [25] Y.D. Zhang, Q. Wang, C.J. Wu, et al., The histogram analysis of diffusion-weighted intravoxel incoherent motion (IVIM) imaging for differentiating the gleason grade of prostate cancer, Eur. Radiol. 25 (2015) 994–1004.