The Ethical Committee of the Provincial Hospital Affiliated to Shandong University (Shandong, China)approved this prospective study and written informed consent was obtained from all of the participants.

The general practice of urologists in China and preference of the majority of patients is to undergo MRI examination prior to biopsy rather than following biopsy. As a routine examination, patients with clinically suspicious PCa (total prostate specific antigen (tPSA), >4.0 ng/ml; free PSA (fPSA)/tPSA, <0.16; positive findings from digital rectal examination) usually underwent a pre-biopsy MRI in the Radiology Department of Provincial Hospital (Shandong University, Shandong, China). From January 2010 to August 2013, 518 patients with clinically suspicious PCa were referred to undergo MRI examination by TRUS-Gbs. Patients with pretreatment, including hormone, radiation and surgery, had been excluded.

All MRI examinations were performed on a 3.0-T system (Magnetom verio; SIEMENS, Munich, Bavaria, Germany) with integrated eight-channel pelvic phased-array surface coils and spine coils for signal reception. All patients were imaged in a supine, head first position.

MR imaging sequences included T2WI, DWI, DCE imaging and MRS, among others. The parameters of sequences are summarized in Table I. In Table I, the sequences with a small field of view (FOV; 20×20 cm²) were used to observe PCa and the invasion of adjacent structures, including organ-confined tumors, extracapsular extension and seminal vesicle invasion. The sequences with large FOVs were used to observe nodal metastases and distant spread. Coronal and/or sagittal enhanced T1-weighted imaging as well as volume interpolated body examination with large FOVs were optional and are not listed in Table I. T2WI was performed with a turbo spin-echo sequence. In order to facilitate positioning, the center of slice groups, slice thickness and FOV used for DWI and DCE imaging were consistent with those of the axial T2WI with small FOVs. MRS data were overlaid on the corresponding axial T2W images. The MRS followed axial, coronal and sagittal T2WI to avoid the non-correspondence between the spectra and the corresponding voxel due to organ movement. DWI was performed using a single-shot echo-planar imaging technique. Apparent diffusion coefficient (ADC) maps were automatically generated from the DWI by the scanner. The MRS was performed using 3D chemical shifting imaging techniques based on a point-resolved spectroscopic sequence with sufficient lipid and water suppression. Eight saturation bands were used to minimize the contamination from adjacent structures of the prostate. Prior to approval for evaluation, a spectroscopist validated the spectra by examining them with regard to the correct positions, signal-to-noise ratio (SNR) >5:1, full width at half maximum (FWHM) ≤15 Hz, a relatively steady baseline and the absence of lipid signals.

Two senior radiologists determined the PCa suspicious regions in consensus using the combined information of multi-parametric MRI. A lesion fulfilling two or more of the following criteria was regarded as a PCa suspicious region: i) An area with homogeneous low-signal-intensity and mass effect in the T2W images; ii) focal hyperintensity in diffusion-weighted images and corresponding hypointensity in ADC maps; iii) a normal choline (Cho) + creatine (Cr)/citrate (Cit) ratio (CC/C) = 0.22±0.12 in the peripheral zone (PZ) and (CC/C) = 0.34±0.14 in the transition zone (TZ) (14) and CC/C >0.34 [the mean ± standard deviation] in PZ and CC/C >0.48 in TZ and iv) a lesion that enhances rapidly and is washout of contrast agent in delay phase.

The median interval time from MRI examination to biopsy was three days (range, 1–12 days). Considering the areas with the minimum ADC or with the maximum CC/C within suspicious regions may represent the most abnormal part with the highest cellularity or metabolism (15,16). These areas were used to target the biopsies to represent the highest grade of the tumor (Figs. 1 and 2). The patients with suspected PCa on MR images underwent TRUS-MR-Dbs besides TRUS-Gbs (12-core) using an ultrasound device (DC-7; Mindray Medical International Limited, Shenzhen, China). The biopsy was performed by an urologist and a radiologist. The radiologist was also involved in the interpretation of the MR images. The steps of biopsy were as follows: In step 1, the 12-core systematic TRUS-Gbs were performed on all patients. In step 2, the area with minimum ADC in the suspicious regions was visually matched with the corresponding location identified by TRUS. One- to two-core biopsy was performed in this area while monitoring by TRUS. This step is known as DWI-Db. In step 3, the area with the maximum CC/C in the suspicious regions was visually matched with the corresponding location identified using TRUS. One- to two-core biopsy was performed in this area while monitoring by TRUS. This step is known as MRS-Db. In step 4, one- to four-core biopsy was performed under the TRUS in the suspicious regions in MR images in addition to the biopsies performed in steps 2 and 3.

If there were no suspicious regions of PCa on MRI, only step 1 was performed. If the minimum ADC and the maximum CC/C appeared in the same site, steps 2 or 3 were omitted. The combination of steps 2–4 is referred to as TRUS-MR-Db in the present study. The highest Gleason grade (HGG) for TRUS-MR-Db was obtained from step 2 and/or step 3. The Gleason score (GS) for TRUS-MR-Db was obtained from step 2, step 3 and step 4. The HGG and GS for TRUS-Gb were obtained from step 1.

TRUS was able to provide dynamic anatomy images of arbitrary sections. The majority of PCa cases were present as hypoechoic masses. However, the sonographic findings of PCa were non-specific. Numerous tumors were iso-echoic on TRUS. To achieve a good spatial consistency between ultrasound scans and MRI sections in performing TRUS-MR-Dbs (step 2, 3 and 4), certain criteria were required to be met for the biopsy: The gland morphology determined using ultrasound axial and sagittal scans being similar to that obtained with the axial and sagittal T2W images, in regard to specific anatomic structures (i.e. seminal vesicles, veromontanum, urethra and a number of hyperplasia-nodules) as landmarks. The biopsy cores were labeled to specify the location of the biopsy and marked on the T2W images at the corresponding sites.

Collectively, 12–18 cores were obtained from each patient. A median of four cores (range, two-six), which contained a median of two cores (range, one-three) in the most abnormal parts, were obtained from the suspicious regions for PCa in each patient.

The sites of suspicious PCa and their number were recorded on MRI for each patient. TRUS-MR-Dbs cores corresponding to each suspicious tumor on MRI were also recorded. The sites of 12 cores of TRUS-Gbs were relatively fixed. In the present study, only the patients who underwent RP were included to predict the aggressiveness of the PCa. The number of tumors was confirmed by RP pathology. For the patients who underwent RP, the urologist, the pathologist and the radiologist confirmed in consensus which cores of TRUS-Gbs and which cores of TRUS-MR-Dbs corresponded with each tumor using the biopsy records, the biopsy/RP pathology and the MR images as references. The biopsy cores corresponding to each tumor in a patient were then confirmed by the two biopsy methods. Only the tumors visible on MR images were included.

The median interval time from biopsy to RP was nine days (2–33 days). Pathological sections of prostate from RP were matched with MR sections based on the level sextant locations by a pathologist and a radiologist. For cancerous biopsy cores and RP specimens, the GS of each tumor were determined by a pathologist with 15 years of experience, by the sum of the primary and secondary Gleason grades. The corresponding HGG of the specimens from the most aberrant regions was identified. The tumor volume was calculated by using the ellipsoid formula (0.52 × length × width × height).

Cross-tabulation analysis was used to describe the biopsy and RP findings (including HGG and GS). For both TRUS-Gbs and TRUS-MR-Dbs, the concordance rates (CRs) with RP were determined for HGG and GS. McNemar tests or χ² analyses with Fisher’s exact tests were performed to determine the differences for CRs with RP between TRUS-MR-Dbs and TRUS-Gbs as well as between DWI-Dbs and MRS-Dbs. The independent samples t-test was used to determine the difference between the volumes of tumors with accurate GS and of tumors with lower GS for TRUS-MR-Dbs. The correlation between tumor volume and HGG was evaluated using Pearson’s correlation. For all statistical analyses, P<0.05 was considered to indicate a statistically significant difference. Statistical analysis was performed with SPSS for Windows, version 17.0 (SPSS, Inc., Chicago, IL, USA).

MRI results were positive in 254 (49.0%) of the 518 patients. TRUS-MR-Db alone detected 165/254 (65.0%) of cancer cases. TRUS-Gb alone detected 190/518 (36.7%) of cancer cases. The overall number of patients with cancer was 196. The median age of the 196 patients was 73 years (range, 51–87 years). The 165 cancer cases detected by TRUS-MR-Db included 159 of the 190 cancer cases, which were detected by TRUS-Gb, and six that were not able to be detected by TRUS-Gb. These six cancer cases had a HGG ≥3. There were 31 cancer cases that were not detectable by TRUS-MR-Db, but by TRUS-Gb. Among these 31 cancer cases, 24 had a HGG of ≤3 and 26 were <0.5 cm in length.

Out of the 165 patients, only 40 patients underwent RP and the quality of MR images and MRS of these patients matched the criteria. The clinical results of 40 patients with 48 tumors (8 patients, 2 tumors/patient; 32 patients, 1 tumor/patient) are presented in Table II. Tables III and IV summarize the biopsy and RP findings of these patients. TRUS-MR-Db demonstrated a higher CR with RP for overall HGG, 89.6% (43/48). For TRUS-Gb, it was 72.9% (35/48) (P=0.008). The CRs with RP for TRUS-MR-Dbs vs. those for TRUS-Gbs were 100% (14/14) vs. 85.7% (12/14; P=0.5) for tumors with HGG of 3 (HGG3); 87.5% (28/32) vs. 68.8% (22/32; P=0.031) for tumors with HGG4 and 50% (1/2) vs. 50% (1/2; P=1) for tumors with HGG5 (Fig. 3A). For biopsies with low grade (HGG3), the positive predictive value (PPV) for TRUS-MR-Dbs representing a true low grade was 77.8% (14/18), whereas for TRUS-Gbs it was 54.5% (12/22; P=0.125). Undergrading of tumors compared with RP was present at 10.4% (5/48) for TRUS-MR-Dbs and 27.1% (13/48) for TRUS-Gb (P=0.008). Undergrading of tumors to HGG<4, which had been graded as HGG4 or HGG5 by RP, occurred at 29.4% (10/34) for TRUS-Gbs and 11.8% (4/34) for TRUS-MR-Dbs (P=0.031). No overgrading was observed for the biopsies.

The HGG CRs with RP determined by DWI-Dbs vs. MRS-Dbs were 77.1% (37/48) vs. 50.0% (24/48; P=0.015) for overall tumors; 80.0% (28/35) vs. 40.0% (14/35; P=0.003) for PZ tumors and 69.2% (9/13) vs. 76.9% (10/13; P=1) for TZ tumors. For DWI, the difference of HGG CRs with RP between PZ and TZ tumors was not significant (P=0.43). For MRS, the difference between PZ and TZ tumors was significant (P=0.023).

A total of 37 (77.1%) and 25 (52.1%; P=0.004) tumors were assigned the same GS as with RP by TRUS-MR-Dbs and TRUS-Gbs, respectively (Fig. 3B). By TRUS-MR-Dbs and TRUS-Gbs, 9 (18.8%) and 23 (47.9%; P<0.001) tumors were assigned a lower GS than that assigned by RP, respectively. Two (4.5%) tumors were assigned a higher GS than that assigned by RP for TRUS-MR-Dbs. The GSs determined by biopsies and RP are presented in Table V. The volume of tumors assigned the same GS as RP for TRUS-MR-Dbs (5.3±4.0 cm³) was significantly smaller than that of the tumors assigned lower GS (9.1±5.3 cm³; P=0.02). The tumor volume demonstrated a positive correlation with the HGG (r=0.396; P=0.005).