Mean diffusivity discriminates between prostate cancer with grade group 1&2 and grade groups equal to or greater than 3

European Journal of Radiology (2016) 1–1801

Contents lists available at ScienceDirect

European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad

Mean diffusivity discriminates between prostate cancer with grade group 1&2 and grade groups equal to or greater than 3 M. Nezzo a,∗ , M.G. Di Trani b , A. Caporale c,d , R. Miano e , A. Mauriello f , P. Bove e , S. Capuani d , G. Manenti a a Department of Diagnostic and Interventional Radiology, Molecular Imaging and Radiotherapy, PTV Foundation, “Tor Vergata” University of Rome, Viale Oxford 81, 00133 Rome, Italy b Physics Department, Sapienza University of Rome, Piazzale Aldo Moro 5, Rome, Italy c Department of Anatomical, Histological, Forensic and Locomotor System Science, Morfogenesis and Tissue Homeostasis, Sapienza University of Rome, Italy d CNR ISC, UOS Roma Sapienza, Physics Department Sapienza University of Rome, Italy e Urology Unit, Department of Experimental Medicine and Surgery, PTV Foundation, “Tor Vergata” University of Rome, Viale Oxford 81, 00133 Rome, Italy f Anatomic Pathology, Department of Biomedicine and Prevention, PTV Foundation, “Tor Vergata” University of Rome, Viale Oxford 81, 00133 Rome, Italy

a r t i c l e

i n f o

Article history: Received 18 May 2016 Received in revised form 28 June 2016 Accepted 1 August 2016 Keywords: Prostate cancer Grade group Magnetic resonance imaging Diffusion tensor imaging MD

a b s t r a c t Purpose: To test the potential ability of mean diffusivity (MD) and fractional anisotropy (FA) in discriminating between PCa of grade group (GG) 1&2, and GGs ≥ 3. Material and methods: Diffusion Tensor Imaging (DTI) experiments at 3T in a cohort of 38 patients with PCa (fifty lesions in total) were performed, by using different diffusion weights (b values) up to 2500 s/mm2 . Gleason score (GS) and GG data were correlated with DTI parameters (MD and FA) estimated in PCa. The relation between DTI measures and GS was tested by the linear correlation analysis (Pearson’s coefficient). One-way analysis of variance to check the statistical significance of the difference between GG 1&2 and GGs 3, 4, 5, ≥3 was used. Results were reported for each of the three b-values ranges: 0–800 s/mm2 , 0–1500 s/mm2 , 0–2500 s/mm2 . Results: A negative correlation was found between MD and GS. The highest linear correlation was observed when the fit was performed with data acquired in the b-values range 0–2500 s/mm2 . MD values were significantly different between GG 1&2 and GG = 3 and between GG 1&2 and GG ≥3. Moreover this difference is better defined when high b values (higher than b = 800 s/mm2 ) are used. The specificity, sensitivity and accuracy in the discrimination between GG 1&2 and GG = 3 were: 90%, 66.7% and 82.4%, respectively when MD was estimated in the b-values range 0–2500 s/mm2 while these values were 85%, 58.3% and 78.4% when MD was estimated in the b-values range 0–800 s/mm2 . Conversely FA did not discriminate between GG 1&2 and GG ≥3, at any investigated b-values range. Conclusion: This study suggests that MD estimation in PCa, obtained from DTI acquired at high b-values, can contribute to the diagnosis and grading of prostate cancer while FA is not a useful parameter for this purpose. © 2016 Elsevier Ireland Ltd. All rights reserved.

1. Introduction Prostatic adenocarcinoma (PCa) is the most frequent cancer among males in Europe [1–4]. As a consequence new diagnostic imaging techniques and reliable diagnostic criteria are highly desirable. The purpose of the present retrospective study was to test the potential ability of diffusion tensor imaging (DTI) derived

∗ Corresponding author. E-mail address: [email protected] (M. Nezzo). http://dx.doi.org/10.1016/j.ejrad.2016.08.001 0720-048X/© 2016 Elsevier Ireland Ltd. All rights reserved.

parameters in discriminating between PCa with grade group (GG) 1&2 (low risk cancer), and GGs ≥3 (intermediate/high risk cancer). According to the European Association of Urology (EAU) guidelines on prostate cancer [3], the standard diagnostic procedure to confirm the presence of PCa is based on the transrectal or transperineal ultrasonography (TRUS) biopsy. However, the chance of missing a PCa by sextant biopsy using a computerized biopsy simulation is about 25% [5]. Furthermore, while TRUS biopsy has a good accuracy, it may have discrepancies with prostatectomy specimens [6]. From prostatic biopsy a Gleason score is assigned to evaluate the aggressiveness of PCa. For men who suffer from aggressive

2

M. Nezzo et al. / European Journal of Radiology (2016) 1–1801

PCa characterized by a Gleason score (GS) ≥4 + 3 radical prostatectomy or irradiation of the prostate gland are indicated. Conversely, patients suffering from a less aggressive (GS ≤ 3 + 4) cancer can be offered less aggressive and invasive treatment [4–6]. As a consequence, the knowledge of the PCa aggressiveness before invasive surgery, is highly desirable to help plan cancer treatments for ensuring all of the tumors treated sparing as much normal tissue as possible. More importantly, a correct differentiation between low risk and intermediate/high risk cancer is desirable to ensure an appropriate management of patients, which can even include focal therapies or active surveillance [7]. In this regard, Epstein et al. [8] have recently proposed a new grading system for PCa to provide a better classification of cancer aggressiveness and thus a more correct prognosis for the patient [8]. In particular they showed [8] a different biochemical recurrencefree progression rate after radical prostatectomy in each assigned grade group (from 1 to 5 according to the Gleason score, GS), highlighting a large difference between PCa with grade group (GG) 1 and 2 (1&2), characterized by GS = 3 + 3 and 3 + 4, respectively compared to GG 3 (characterized by GS = 4 + 3) and greater than 3 (characterized by GS ≥8). In particular, a significant difference between GG 1&2 and GG 3 in terms of prognosis and recurrence after treatment, has been highlighted. During the last years, multiparametric MRI has been widely used to detect tumor lesions and assess their aggressiveness [9–13]. Importantly, diffusion weighted imaging (DWI) has been indicated as a “dominant” sequence for PCa detection in the peripheral zone (PZ) and it can be helpful for PCa diagnosis in the transition zone (TZ) [14]. DWI provides information about the behavior of water molecules diffusion in tissue that is influenced by tissue topology and microstructures that impede and hinder water motion within tissues. In prostatic tissue, the branching ductal and the acini structure of the normal prostate compared with the highly restricted intracellular and interstitial spaces encountered in PCa produces a substantial differential in DWI image contrast. Recently, some authors have shown that high b-value DWI images (e.g. with b-values greater than 800 s/mm2 ) allow increased delineation of PCa [15,16]. Some other, underlined that DTI could provide a more accurate investigation of the prostatic tissue than that furnished by DWI [17–25]. From DTI measurements, it is possible to derive the mean diffusivity (MD) of water in tissues and various measures of its diffusion anisotropy, such as the fractional anisotropy (FA). In particular, MD was significantly lower in PCa compared to benign prostate tissue [21–23] and a significant correlation between DTI measurements and GS in PCa was found [24,25]. Aim of the present study was to test MD and FA potential ability in discriminating between PCa with GG 1&2, and GGs equal to or greater than 3. Toward this goal, we performed DTI experiments at 3T with b values up to 2500 s/mm2 in a cohort of 38 patients with PCa. We correlated the anamnestic and histological patients’ data with DTI parameters measured by using monoexponential fits performed with data obtained at different b-values ranges: (a) from 0 to 800 s/mm2 , (b) 0 to 1500 s/mm2 and (c) 0 to 2500 s/mm2 . MD and FA results obtained in PCa belonging to different GGs were compared by using statistical tests. Finally, the relation between the DTI parameters and the GS was investigated in the three above mentioned b-values ranges.

2. Materials and methods 2.1. Patient cohort Between February and November 2015, a total of 68 subjects with a possible diagnosis of PCa were scanned prior to

their first biopsy. The mean age of the patients was 70.8 (age range: 48–86 years) and the averaged PSA was 10,1 ng/mL (PSA range 3,7–26,2 ng/mL). Informed consent was obtained from each patient prior to the MRI examination. The study was approved by the Local Ethics Committee. The biopsy, performed after the MRI examination, indicated 38 patients with GS equal or higher than 3 + 3 (Table 1) and 30 subjects with benign histopathological findings. Seventy-eight image slices were used to evaluate DTI parameters in PCa areas. Fifty lesions in total (16 lesions in TZ and 34 lesions in PZ) were investigated (Table 1). Prostatic tissue of the 30 subjects with negative biopsies was also analyzed by considering sixty image slices to evaluate DTI parameters in benign prostatic tissue.

2.2. MRI All the examinations were performed using a 3T clinical MRI system (Intera Achieva, Philips Medical Systems, The Netherlands) equipped with high performance gradients with maximum strength of 80 mT/m and a slew rate of 200 mT/m/ms. For all of the examinations, six-channel phased array SENSE torso coil was used. For each patient, the protocol included high spatial resolution T2-weighted turbo spin echo (TSE) and DTI with echo-planar imaging (EPI). T2-weighted TSE images (repetition time (TR) = 3957, echo time (TE) = 150, turbo factor 21, field of view (FOV) = 150 × 130 mm, slice thickness (STK) = 3 mm, gap = 0, acquisition matrix 256 × 178, reconstruction matrix = 512 × 512, number of averaged scans (NSA) = 6, flip angle = 90◦ ) were obtained for all subjects including the entire gland in the axial plane. DTI protocol was performed with a single-shot EPI sequence (TR = 3000, TE = 67, FOV = 150 × 130 × 70 mm3 , acquisition matrix = 64 × 52, reconstruction matrix 96 × 96, STK = 3 mm gap = 0, NSA = 4), by using 7 b-values (0, 500, 800, 1000, 1500, 2000, 2500 s/mm2 ) and 6 non co-planar gradient diffusion directions. Spectral Attenuated Inversion Recovery (SPAIR) fat suppression with 200 Hz frequency offset was used after a B0 homogeneity optimization by using high order shim routine. The duration of the entire protocol was approximately 12 min of which the length of the DTI protocol is about 9 min. T2-weighted images (T2WIs) were used as anatomical and morphological reference to determine biopsy zones and as DTI reference image.

2.3. Biopsy Biopsies were performed to all the patients in a period of 1 day–2 weeks after the MRI examination by expert urologists. T2WIs were used to contour and record lesion locations (Watson Elementary® ). We performed a targeted MR/ultrasound fusion biopsy (BiopSee® , Medcom, Darmstadt, Germany) obtaining from 2 to 4 biopsy cores from the targets, followed in the same session by a 12 cores transperineal biopsy (sextant and laterally directed biopsies at base, mid-gland and apex). The targeted MR/ultrasound fusion biopsy and the standard 12 cores transperineal biopsy were performed by different physicians; the one who performed the 12 cores transperineal biopsy was unaware of the MR findings. Histopathological examination was performed and reviewed for each specimen on the basis of the recommendations arising from the “consensus conference ISUP 2014” [26].

M. Nezzo et al. / European Journal of Radiology (2016) 1–1801

3

Table 1 Details of PCa lesions evaluated. GG

GS

Patients’ number

Number of lesions

Number of slices

Number of PZ lesions

Number of TZ lesions

1 2 3 4 5* Total

3+3 3+4 4+3 4+4 4 + 5; 5 + 4

12 8 8 6 4 38

16 13 10 7 4 50

20 17 15 16 10 78

10 8 8 5 3 34

6 5 2 2 1 16

*

In GG = 5, two patients had GS = 4 + 5 and two patients had GS = 5 + 4.

Table 2 MD and FA in benign tissue and PCa evaluated at different b-values range. MD (10−3 mm2 /s)

b (s/mm2 )

*

benign

1.60 ± 0.27 1.20 ± 0.19 0.80 ± 0.11

0–800 0–1500 0–2500

P

FA

PCa −20

0.74 ± 0.15 0.60 ± 0.09 0.46 ± 0.07

10 10−22 10−21

benign*

PCa

P

0.22 ± 0.09 0.22 ± 0.07 0.21 ± 0.06

0.30 ± 0.11 0.27 ± 0.10 0.26 ± 0.08

10−4 0.0017 0.0002

P = P-value of the t-test. * prostatic tissue of subjects with benign histopathological findings.

2.4. Diffusion model To obtain MD and FA in each image voxel, the first step was to estimate the apparent diffusivity maps, ADCi for each diffusion gradient direction i, where i = 1,2,3,4,5,6: S(bi ) = S(0)exp(−ADCi ∗bi )

(1)

where S(bi ) is the DW signal, the index i corresponds to a unique gradient direction i, S(0) is the b = 0 signal and bi is the amount of diffusion weighting along i direction. Subsequently, the six independent elements of the diffusion tensor (Dxx , Dyy , Dzz , Dxy = Dyx , Dxz = Dzx , and Dyz = Dzy ) were estimated from the apparent diffusivities using multiple linear least squares methods [27,28]. Once obtained all the elements of the diffusion tensor, its eigenvalues D1 , D2 and D3 are achieved through tensor diagonalization. MD is obtained as eigenvalues average: MD = 1/3(D1 + D2 + D3 )

mono-exponential fitting procedure with the DTIFIT tool by using data acquired with: (a) the whole range of b-values up to 800 s/mm2 , (b) the whole range of b-values up to 1500 s/mm2 , (c) the whole range of b-values, up to b = 2500 s/mm2 . Regions of interest (ROIs) in TZ and PZ (in every image of the patients with benign and malign histopathological findings) were manually drawn on the b0-images by referring to T2WIs. The hypointense foci on T2WIs and restricted diffusion areas in DWIs were carefully examined, and the pathological areas were determined by two experienced uro-radiologists along with a physicist expert in the images treatment, blinded to the patient information (MN, GM and SC). Calcifications, necrosis, and neurovascular bundles were excluded. Masks of the selected ROIs were obtained by using FSL 5.0 and applied to the MD and FA maps. Mean and standard deviation (SD) of the aforementioned DTI parameters were computed with a custom-made MATLAB script (MATLAB R2012b, The Mathworks, Natick, MA) within the selected ROIs.

(2)

and it is independent of any tissue directionality, while FA, which measures the degree of diffusion anisotropy reflecting the degree of cellular structures alignment, is given by:

 

FA =

3 2

2.6. Statistical analysis (D1 − MD)2 + (D2 − MD)2 + (D3 − MD)2



D1 2 + D2 2 + D3 2

(3)

where FA varies from 0 to 1, with FA = 0 representing isotropic diffusion (D1 = D2 = D3) and FA = 1 (D1 = 1, D2 = D3 = 0) representing 100% directional preference along the major eigenvector D1 . 2.5. Data analysis Diffusion weighted images (DWIs) were co-registered and corrected for motion artifacts. The image pre-processing was performed with FSL 5.0 software (FMRIB Software Library v5.0, FMRIB, Oxford, UK) [29]. T2WIs were co-registered to the first acquired volume of the diffusion experiment, corresponding to the one without diffusion weight (b = 0 s/mm2 , b0-volume), in order to provide a reference space to better delineate lesions boundaries. The DWIs were realigned to the b0-volume through a rigid-body transformation with 6 degrees of freedom and a least-squares cost function, with the FMRIB’s Linear Image Registration Tool (FLIRT). The same software was used to extract the diffusion tensor and the rotationally invariant parameters MD and FA, computed performing a

A paired t-test was performed to assess statistical significance of MD and FA parameters between PCa in PZ and TZ, and between benign and PCa prostatic tissue in each of the three considered bvalues ranges (a), (b) and (c). A statistical analysis was performed by using one-way analysis of variance (ANOVA) test to check the statistical significance of the difference between GG 1&2 and GGs 3, 4, 5, ≥3 and between benign prostatic tissue and GG 1&2. A p-value ≤0.05 was considered as statistically significant for both paired t-test and ANOVA test. Receiver Operator Characteristic (ROC) analysis was performed to estimate the different diagnostic performance for discriminating: GG 1&2 and GG 3; GG 1&2 and GGs ≥3 by using MD and FA parameters and to define the most reliable cutoff points to differentiate GG 1&2 and GGs ≥3 in the three different dataset (a), (b) and (c). Sensibility, specificity, cut-off values, ROC area under curve (AUC) and accuracy were calculated. Finally, the linear relation between DTI measures and GS was tested by the linear correlation analysis (Pearson’s coefficient). Pvalues <0.05 were considered statistically significant.

4

M. Nezzo et al. / European Journal of Radiology (2016) 1–1801

Table 3 MD vs GS Pearson’s correlation-test results. b range (s/mm2 )

p-value

r

0–800 0–1500 0–2500

10−3 10−4 10−4

−0.38 −0.43 −0.51

3. Results 3.1. DTI discrimination between benign and PCa tissue Both MD and FA didn’t discriminate between PCa in PZ and TZ in each of the three considered b-values ranges (a) 0–800 s/mm2 , (b) 0–1500 s/mm2 and (c) 0–2500 s/mm2 . Therefore, the mean ± SD values of MD and FA evaluated in PCa lesions are listed in Table 2, along with MD and FA (PZ and TZ averaged values) obtained from patients with benign histopathological findings, for the three considered b-values ranges. 3.2. DTI correlation with GS A moderate inverse correlation between MD and GS was found (Fig. 1, Table 3). The highest linear correlation was observed when the fit was performed with data acquired in the b-values range 0–2500 s/mm2 . No correlation between FA values and GS was found. 3.3. DTI discrimination between different GG In Fig. 2 an example of MD and FA maps is displayed for four patients characterized by a PCa of GG 1 (GS = 3 + 3), GG 2 (GS = 3 + 4), GG 3 (GS = 4 + 3) and GG 4 (GS = 4 + 4). FA didn’t discriminate between GG 1&2 and GG ≥3, at any investigated b-values range. MD significantly discriminated between GG 1&2 and GG 3 in all the considered b-values ranges (a), (b) and (c) (Table 4). Moreover, MD didn’t discriminate between GG 1 and GG 2 when dataset in b values range (a) and (c) were used (Table 4). Conversely, MD discriminated between GG 1 and GG 2 only when dataset in the b values range 0–1500 s/mm2 was used. On the other hand, MD values were significantly higher in GG 1&2 compared to the other GGs ≥ 3 without great overlap in MD values (Fig. 1, Table 4). MD significantly differentiated between PCa of GG 1&2 and PCa of GG 4, GG 1&2 and GG5 (Table 4). Moreover, MD significantly differentiated between benign prostatic tissue and PCa of GG 1&2 (Table 4). To differentiate GG 1&2 and GG ≥3 with MD, the best sensitivity is obtained by using dataset in b-values range 0–2500 s/mm2 while the best specificity is obtainable at low b-values range (Table 5). The MD cut-off values to differentiate GG 1&2 and GGs ≥3, the accuracy and the AUC in the three b-values ranges are reported in Table 5, while in Fig. 3a the characteristic ROC curves are displayed to highlight MD performance in discriminating between GG 1&2 and GGs ≥3 in the three b-values ranges. Moreover, to differentiate GG 1&2 and GG 3 with MD, the best sensitivity and specificity is obtained by using dataset in b-values ranges 0–2500 s/mm2 (Table 6). The MD cut-off values to differentiate GG 1&2 and GG 3, the accuracy and the AUC in the three b-values ranges are reported in Table 6, while in Fig. 3b the characteristic ROC curves are displayed to highlight MD performance in discriminate between GG 1&2 and GG 3 in the three b-values ranges. 4. Discussion Recently, the International Society of Urological Pathology [8] underlined that there is a significant difference in terms of

Fig. 1. Graphs of correlation between MD and GS in the three different b-values ranges used to acquire DTI data. The bars indicates the standard deviation of MD values.

M. Nezzo et al. / European Journal of Radiology (2016) 1–1801

5

Fig. 2. Example of MD and FA maps obtained in four patients characterized by a PCa of GG 1 (GS = 3 + 3), GG 2 (GS = 3 + 4), GG 3 (GS = 4 + 3) and GG 4 (GS = 4 + 4). White arrows indicate the PCa zone. T2-w and DTI maps representative of low risk cancer PCa are displayed in the up red box. T2-w and DTI maps representative of intermediate/high risk cancer PCa are displayed in the blue box. Please note PCa area is really well delineates in MD maps but it is scarcely recognizable in the FA maps.

prognosis and recurrence after treatment between GG 1&2 and GGs ≥3. Indeed, the GGs 3 to 5 have a higher recurrence rate after treatment and more frequent metastases [8] while GG 1&2 is classified as low risk cancer.

Therefore, a reliable classification of the PCa aggressiveness is crucial to choose the correct patients’ management. A reliable diagnostic method of imaging to differentiate between low risk PCa, characterized by GG = 1&2 and intermediate/high risk cancer,

6

M. Nezzo et al. / European Journal of Radiology (2016) 1–1801

Table 4 P-values of ANOVA-test performed to compare MD between different GGs. b (s/mm2 )

GG = 1&2 vs benign tissue*

GG = 1&2 vs GG ≥ 3

GG = 1 vs GG = 2

GG = 1&2 vs GG = 3

GG = 1&2 vs GG = 4

GG = 1&2 vs GG = 5

0–800 0–1500 0–2500

10−15 10−18 10−20

10−3 10−5 10−5

NS 0.03 0.48

0.03 0.01 0.01

0.04 10−3 10−4

3 × 10−4 10−4 10−4

GG = Group Grade. NS = No significant. * prostatic tissue of subjects with benign histopathological findings. Table 5 ROC curve analysis to evaluate MD potential to discriminate of GG = 1&2 and GG ≥ 3. b (s/mm2 )

0–800

0–1500

0–2500

Sensitivity (%) Specificity (%) AUC Accuracy (%) Cut-off (10−3 mm2 /s)

63.2 85.0 0.74 ± 0.06 75.0 0.70

81.6 75.0 0.81 ± 0.05 78.2 0.59

84.2 70.0 0.83 ± 0.05 76.9 0.46

GG = Group Grade.

Table 6 ROC curve analysis to evaluate MD potential to discriminate GG = 1&2 and GG = 3. b (s/mm2 )

0–800

0–1500

0–2500

Sensitivity (%) Specificity (%) AUC Accuracy (%) Cut-off (10- 3 mm2 /s)

58.3 85.0 0.70 ± 0.09 78.4 0.70

66.7 75.0 0.72 ± 0.09 74.5 0.59

66.7 90.0 0.81 ± 0.08 82.4 0.42

GG = Group Grade.

Fig. 3. ROC curves showing MD potential to discriminate GG = 1&2 and GG ≥ 3 (A), GG = 1&2 and GG = 3 (B) as a function of the b-value range selected to acquire DWI images.

characterized by GG ≥3 is desirable to perform plan cancer treatments optimized for ensuring all of the tumors treated sparing as much normal tissue as possible. Our results show that DTI protocols based on MD measurement of PCa can help to achieve this goal. We found a significant difference in MD values between GG 1&2 and GG 3 that can be explained by different tumor architecture among groups. In particular, a carcinoma with GG 1&2 (GS = 3 + 3 and GS = 3 + 4) is characterized by a greater amount of well formed glands, where the water diffusion is fast, i.e it is poorly hindered, compared to a GG 3 which is predominantly formed by poorlyformed/fused/cribriform glands showing a great amount of stromal and connective tissue [8]. As water diffusion in stromal and

connective tissue is a slow diffusion, i.e. it is more hindered and obstructed than in glands and ducts, the presence of poorlyformed/fused/cribriform glands with the consequent reduction of gland and ducts lumens in group grade 3 lead to a lower MD values compared to GG 1&2. As regards GGs >3, MD of GG 4 and 5 showed significantly lower values compared to GG 1&2 and GG 3, due to further reduction up to the disappearance of gland and ducts lumens and the important presence of fibro-muscular stroma and undifferentiated cells where water diffusion is strongly obstructed and restricted, showing an even slower diffusion. Moreover, to discriminate between GG 1&2 and GG 3 with the best sensitivity and specificity, our study suggests to estimate MD from dataset acquired by using b values range 0–2500 s/mm2 . In our opinion this result indicate that non-Gaussian diffusion models [30], performed at high b-values, should be investigated to improve the sensitivity and specificity of diffusion imaging techniques in discriminating between low and intermediate/high risk cancer in prostatic tissue. Finally, by considering the correlation between MD and GS, our results are in agreement with recent investigations performed by Li et al. [24] and Uribe et al. [25]. However, in contrast with Li et al. [24] and according to Uribe et al. [25], the results of our study suggest that FA does not contribute significantly to the diagnostic capabilities of DTI in prostate cancer. Indeed we found no significant correlation between FA and GS. Moreover FA wasn’t significantly different between GG 1&2 and GG ≥3. In this regard, it should be noted that our and Uribe FA results were obtained by acquiring DWI with only 6 gradient diffusion directions, while Li et al. used 32 gradient diffusion directions. Because the reliability of FA results can depends on the number of the gradient diffusion directions used to acquire DWI, the potential diagnostic role of FA should be further investigated. Differently from previous papers [24,25] that explore the potential of DTI to discriminate between different GG, we performed

M. Nezzo et al. / European Journal of Radiology (2016) 1–1801

DTI investigations in three different b-values ranges: 0–800 s/mm2 , 0–1500 s/mm2 , 0–2500 s/mm2 , showing a dependence of MD diagnostic potential on b-value range. Moreover, ROC curves of MD values in PCa were performed to define the most reliable cutoff point to differentiate between GG 1&2 and GG ≥3, as a function of the b-values range. The b-value determines the degree of signal attenuation due to diffusion, allowing the selection of different diffusion behaviors (fast bulk diffusion, slow restricted diffusion) characterized by different MD values of water diffusing in heterogeneous tissue. Fast bulk diffusion, that is detected with low b-values is characterized by higher MD compared to slow diffusion that is quantified by using higher b-values. Therefore our results suggest that MD values quantifying slow diffusion better discriminate between low and intermediate/high risk cancer compared to MD associated to fast bulk diffusion. There were three main limitations to this study. First, the number of analyzed patients was limited. In particular we analyzed a low number of patients belonging to GG 5. Second, the GG 4 includes three histological architectures of cancer with different GS: 4 + 4, 3 + 5 and 5 + 3; in patients with GG 4 examined in our study, no one was found to have a GS = 5 + 3 and a GS = 3 + 5. In this last case MD value could be similar to GG 1&2 because of its predominant well-formed glands associated with a lesser component of lacking glands. As a consequence further studies are necessary to address this issue. Third, we correlated our DTI data with the histological results of the TRUS-MRI fusion-biopsy, not with surgical specimens of the whole prostate. In this regard, a recent work has highlighted that the results of the MR-ultrasound fusion biopsy compared to those of prostatectomy specimens show a correlation of 90% for the evaluation of the primary Gleason grade, 59% for the secondary Gleason grade and 67% as regards total GS with an underestimation of the total GS in 29% of cases [31].

5. Conclusions In conclusion, although the number of patients evaluated was small, our results highlight that there is a significant difference of the MD values between GG 1&2 and GG ≥3 and this difference is more evident when high b-values (higher than b = 800 s/mm2 ) are used. Therefore our study strongly suggests that MD estimation in PCa can contribute to the diagnosis and grading of prostate cancer. This quantitative data, if confirmed with an higher number of patients, could be of fundamental importance to reduce the number of biopsies and avoid overtreatment.

Conflict of interest No conflict of interest.

References [1] M. Arnold, H.E. Karim-Kos, J.W. Coebergh, G. Byrnes, A. Antilla, J. Ferlay, A.G. Renehan, D. Forman, I. Soerjomataram, Recent trends in incidence of five common cancers in 26 European countries since 1988: analysis of the European Cancer Registry database, Eur. J. Cancer 51 (9) (2013) 1164–1187. [2] M.J. Roobol, R. Kranse, C.H. Bangma, A.G.J.L.H. Van Leenders, B.G. Blijenberg, R.H.N. Van Schaik, W.J. Kirkels, S.J. Otto, T.H. Van Der Kwast, H.J. De Koning, F.H. Schröder, Screening for prostate cancer: results of the Rotterdam section of the European randomized study of screening for prostate cancer, Eur. Urol. 64 (4) (2013) 530–539. [3] N. Mottet, P. Bastian, J. Bellmunt, R. van den Bergh, M. Bolla, N. van Casteren, P. Cornford, S. Joniau, V. Matveev, T. van der Kwast, H. van der Poel, O. Rouvière, T. Wiegel, Prostate Cancer (2014). [4] P.C. Walsh, T.L. DeWeese, M.A. Eisenberger, Clinical practice. Localized prostate cancer, N. Engl. J. Med. 357 (26) (2007) 2696–2705.

7

[5] F. Daneshgari, G.D. Taylor, G.J. Miller, E.D. Crawford, Computer simulation of the probability of detecting low volume carcinoma of the prostate with six random systematic core biopsies, Urology 45 (4) (1995) 604–609. [6] S.W. Fine, J.I. Epstein, A contemporary study correlating prostate needle biopsy and radical prostatectomy gleason score, J. Urol. 179 (April) (2008) 1335–1339. [7] H.U. Ahmed, O. Akin, J.A. Coleman, S. Crane, M. Emberton, L. Goldenberg, H. Hricak, M.W. Kattan, J. Kurhanewicz, C.M. Moore, C. Parker, T.J. Polascik, P. Scardino, N. Van As, A. Villers, Transatlantic Consensus Group on active surveillance and focal therapy for prostate cancer, BJU Int. 109 (11) (2012) 1636–1647. [8] J.I. Epstein, L. Egevad, M.B. Amin, B. Delahunt, J.R. Srigley, P.A. Humphrey, G. Committee, The 2014 International Society of Urological Pathology (ISUP) Consensus Conference on Gleason Grading of Prostatic Carcinoma Definition of Grading Patterns and Proposal for a New Grading System, vol. 00 no. 00 (2015) pp. 1–9. [9] J. Richenberg, R. Clements, S. Verma, G. Villeirs, O. Rouviere, V. Logager, J.O. Barentsz, J.J. Fütterer, P. Choyke, ESUR prostate MR guidelines 2012, Eur. Radiol. 22 (4) (2012) 746–757. [10] F. Russo, D. Regge, E. Armando, V. Giannini, A. Vignati, S. Mazzetti, M. Manfredi, E. Bollito, L. Correale, F. Porpiglia, Detection of prostate cancer index lesions with multiparametric MRI (mp-MRI) using whole-mount histological sections as the reference standard, BJU Int. (2015). [11] O.F. Donati, Y. Mazaheri, A. Afaq, H. Vargas, J. Zheng, C.S. Moskowitz, H. Hricak, O. Akin, Prostate cancer aggressiveness: assessment with whole-lesion histogram analysis of the apparent diffusion coefficient, Radiology 271 (1) (2014) 143–152. [12] E. Cho, D.J. Chung, D.M. Yeo, D. Sohn, Y. Son, T. Kim, S.-T. Hahn, Optimal cut-off value of perfusion parameters for diagnosing prostate cancer and for assessing aggressiveness associated with Gleason score, Clin. Imaging (2015). [13] K.M. Selnæs, I.S. Gribbestad, H. Bertilsson, A. Wright, A. Angelsen, A. Heerschap, M.B. Tessem, Spatially matched in vivo and ex vivo MR metabolic profiles of prostate cancer – investigation of a correlation with Gleason score, NMR Biomed. 26 (5) (2013) 600–606. [14] J.C. Weinreb, J.O. Barentsz, P.L. Choyke, F. Cornud, M.A. Haider, K.J. Macura, D. Margolis, M.D. Schnall, F. Shtern, C.M. Tempany, H.C. Thoeny, S. Verma, PI-RADS prostate imaging – reporting and data system: 2015, version 2, Eur. Urol. (2015). [15] A.B. Rosenkrantz, N. Hindman, R.P. Lim, K. Das, J.S. Babb, T.C. Mussi, S.S. Taneja, Diffusion-weighted imaging of the prostate: comparison of b1000 and b2000 image sets for index lesion detection, J. Magn. Reson. Imaging 38 (3) (2013) 694–700. [16] A.B. Rosenkrantz, S.S. Taneja, Radiologist, be aware: ten pitfalls that confound the interpretation of multiparametric prostate MRI, Am. J. Roentgenol. 202 (1) (2014) 109–120. [17] S. Sinha, U. Sinha, In vivo diffusion tensor imaging of the human prostate, Magn. Reson. Med. 52 (3) (2004) 530–537. [18] G. Manenti, M. Carlani, S. Mancino, V. Colangelo, M. Di Roma, E. Squillaci, G. Simonetti, Diffusion tensor magnetic resonance imaging of prostate cancer, Invest. Radiol. 42 (6) (2007) 412–419. [19] J. Xu, P. Humphrey, A.S. Kibel, A.Z. Snyder, V.R. Narra, J.J.H. Ackerman, S.-K. Song, Magnetic resonance diffusion characteristics of histologically defined prostate cancer in humans, Magn. Reson. Med. 61 (4) (2009) 842–850. [20] P. Kozlowski, S.D. Chang, R. Meng, B. Mädler, R. Bell, E.C. Jones, S.L. Goldenberg, Combined prostate diffusion tensor imaging and dynamic contrast enhanced MRI at 3T – quantitative correlation with biopsy, Magn. Reson. Imaging 28 (5) (2010) 621–628. [21] B. Gürses, N. Tasdelen, F. Yencilek, N.O. Kilickesmez, T. Alp, Z. Firat, M.S. Albayrak, A.M. Uluˇg, N. Gürmen, Diagnostic utility of DTI in prostate cancer, Eur. J. Radiol. 79 (2011) 172–176. [22] C. Reischauer, B.J. Wilm, J.M. Froehlich, A. Gutzeit, L. Prikler, R. Gablinger, P. Boesiger, K.U. Wentz, High-resolution diffusion tensor imaging of prostate cancer using a reduced FOV technique, Eur. J. Radiol. 80 (2) (2011) e34–e41. [23] M. Moradi, S.E. Salcudean, S.D. Chang, E.C. Jones, N. Buchan, R.G. Casey, S.L. Goldenberg, P. Kozlowski, Multiparametric MRI maps for detection and grading of dominant prostate tumors, J. Magn. Reson. Imaging 35 (6) (2012) 1403–1413. [24] L. Li, D.J. a Margolis, M. Deng, J. Cai, L. Yuan, Z. Feng, X. Min, Z. Hu, D. Hu, J. Liu, L. Wang, Correlation of gleason scores with magnetic resonance diffusion tensor imaging in peripheral zone prostate cancer, J. Magn. Reson. Imaging (2014) 1–8. [25] C.F. Uribe, E.C. Jones, S.D. Chang, S.L. Goldenberg, S. a. Reinsberg, P. Kozlowski, In vivo 3T and ex vivo 7T Diffusion Tensor Imaging of prostate cancer: correlation with histology, Magn. Reson. Imaging 33 (5) (2015) 577–583. [26] J.I. Epstein, L. Egevad, M.B. Amin, B. Delahunt, J.R. Srigley, P.A. Humphrey, The grading committee, the 2014 international society of urological pathology (ISUP) consensus conference on gleason grading of prostatic carcinoma: definition of grading patterns and proposal for a new grading system, Am. J. Surg. Pathol. 40 (2) (2016) 244–252. [27] P.J. Basser, J. Mattiello, D. LeBihan, Estimation of the effective self-diffusion tensor from the NMR spin echo, J. Magn. Reson. B 103 (3) (1994) 247–254. [28] K.M. Hasan, D.L. Parker, A.L. Alexander, Comparison of gradient encoding schemes for diffusion-tensor MRI, J. Magn. Reson. Imaging 13 (5) (2001) 769–780.

8

M. Nezzo et al. / European Journal of Radiology (2016) 1–1801

[29] M.W. Woolrich, S. Jbabdi, B. Patenaude, M. Chappell, S. Makni, T. Behrens, C. Beckmann, M. Jenkinson, S.M. Smith, Bayesian analysis of neuroimaging data in FSL, Neuroimage 45 (1 Suppl) (2009). [30] S. De Santis, A. Gabrielli, M. Palombo, B. Maraviglia, S. Capuani, Non-Gaussian diffusion imaging: a brief practical review, Magn. Reson. Imaging 29 (10) (2011) 1410–1416.

[31] C. Lanz, F. Cornud, F. Beuvon, A. Lefèvre, P. Legmann, M. Zerbib, N.B. Delongchamps, Gleason score determination with transrectal ultrasound-magnetic resonance imaging fusion guided prostate biopsies – are we gaining in accuracy? J. Urol. 195 (1) (2016) 88–93.