To evaluate the Prostate Imaging Reporting and Data System (PIRADS) in multiparametric magnetic resonance imaging (mpMRI) based on single cores and single-core histology.
To calculate positive (PPV) and negative predictive values (NPV) of different modalities of mpMRI.
Patients and Methods
We performed MRI-targeted transrectal ultrasound-guided perineal prostate biopsies on 50 patients (mean age 66 years, mean PSA level of 9.9 ng/mL) with suspicion of prostate cancer. The biopsy trajectories of every core taken were documented in three dimensions (3D) in a 3D-prostate model.
Every core was evaluated separately for prostate cancer and the performed biopsy trajectories were projected on mpMRI images.
PIRADS scores of 1177 cores were then assessed by a histology ‘blinded’ uro-radiologist in T2-weighted (T2W), dynamic contrast-enhanced (DCE), diffusion-weighted imaging (DWI) and magnetic resonance spectroscopy (MRS).
The PIRADS score was significantly higher in cores positive for cancer than in negative cores.
There was a significant correlation between the PIRADS score and histopathology for every modality.
Receiver operating characteristic (ROC) analysis showed excellent specificity for T2W (90% peripheral zone/97% transition zone) and DWI (98%/97%) images regardless of the prostate region observed. These numbers decreased for DCE (80%/93%) and MRS (76%/83%).
All modalities had NPVs of 99%, if a PIRADS score threshold of 2 (for T2W, DCE, and MRS) or 3 (for DWI) was used. However, PPVs were low.
Our results show that PIRADS scoring is feasible for clinical routine and allows standardised reporting.
PIRADS can be used as a decision-support system for targeting of suspicious lesions.
mpMRI has a high NPV for prostate cancer and, thus, might be a valuable tool in the initial diagnostic evaluation.
Prostate cancer is the most common solid tumour and the third leading cause of male cancer deaths in developed countries . The standard diagnostic pathway consists of systematic ultrasound-guided random biopsies of the prostatic gland. However, a significant number of transrectal biopsies are negative for cancer and provide imprecise results [2-4]. Additionally, Gleason scores between biopsy and prostatectomy specimen show clinically relevant upgrading in 30–40% of cases . Furthermore, TRUS-guided transrectal biopsies have limitations in sampling the anterior regions of the prostate. Evaluation of radical prostatectomy specimens recently showed that a significant number of ventral tumours were missed by transrectal biopsies .
According to the latest evidence, mpMRI can exclude clinically significant prostate cancer with negative predictive values (NPVs) of around 80–90% . mpMRI allows for an improved depiction of tumour foci by its high soft tissue contrast-resolution . To combine those advantages with the practicability of TRUS-guided prostate biopsies, several fusion systems are available on the market, which have shown improved detection rates for prostate cancer .
One problem of image-fusion guided prostate biopsies is the objective and reproducible reporting of MRI findings to the urologist. To solve this limitation, the European Society of Urogenital Radiology (ESUR) recently described the Prostate Imaging Reporting and Data System (PIRADS) for semi-quantitative, standardised reporting of prostate mpMRI .
The aim of the present study was to validate the PIRADS scoring system using transperineal MRI/TRUS-fusion biopsies from patients with suspected prostate cancer on a core-by-core basis with histology as a standard of reference.
Patients and Methods
Our retrospective single-institution study was performed according to the standards of our local ethics board and informed consent was obtained from each patient. In all, 50 consecutive patients with a median (range) age of 66 (48–78) years and a mean (range) PSA level of 9.9 (2.7–49) ng/mL with suspicion of prostate cancer due to elevated PSA level and/or abnormal DRE where included in the study (Table 1). Subsequently MRI-targeted TRUS-guided perineal prostate biopsies were taken as described recently . A median (range) of 24 (13–32) cores with 4 (1–6) targeted and 20 (12–26) systematic cores were taken per patient. In all, 55% (27/50) of patients underwent a first prostate biopsy and 45% (23/50) had had at least one prior prostate biopsy. Three patients had been diagnosed with low-risk prostate cancer by previous biopsies and were under active surveillance.
Table 1. The patients' characteristics
Positive for prostate cancer
Negative for prostate cancer
29 (26 with intermediate- or high-risk disease )
Median (range) age, years
Mean (range) PSA level, ng/mL
Mean (range) prostate volume, mL
All patients underwent mpMRI on a 3-T MRI (Magnetom TRIO, Siemens Medical Solutions, Erlangen, Germany) with a phased-array surface body coil. Detailed parameters of the sequences used are given in Table 2.
Table 2. Sequence parameters of mpMRI
TSE, turbo spin-echo; TWIST, time-resolved angiography with stochastic trajectories; csi, chemical shift imaging; SE, spin echo; epi, echo planar imaging; 2D, two dimensional; 3D, three dimensional.
Repetition time ms/echo time ms
Flip angle, °
Echo train length/epi factor
0, 50, 100, 150, 200, 250, 800
Section thickness, mm
Field of view, mm
1.1 × 1.0
0.8 × 0.7
2.2 × 2.2
1.6 × 1.6
6.0 × 6.0
Acquisition time, min
Every core was embedded separately and evaluated histologically for the presence and grade of prostate cancer. The numbers of the cores were assigned during intervention and the biopsy trajectories of every core taken were documented in a three dimensional (3D)-prostate model . For further analysis, patients were divided in to high-, intermediate- or low-risk disease according to the National Comprehensive Cancer Network (NCCN) classification .
For imaging analyses T2-weighted (T2W), dynamic contrast-enhanced (DCE), diffusion-weighted imaging (DWI) and magnetic resonance spectroscopy (MRS) images were evaluated in consensus by two radiologists with >10 years and 2 years of experience in genitourinary imaging, respectively.
Next, we retrospectively projected the performed biopsy trajectories on mpMRI images (Fig. 1). PIRADS scores for 1177 cores were then assessed by a histology ‘blinded’ uro-radiologist in T2W, DCE, DWI and MRS separately. Only the intraprostatic portion of the cores and imaged prostate parenchyma without partial volume effects were assigned with a PIRADS score. An overall PIRADS score was not assigned.
We correlated retrospective PIRADS scores to clinical data, i.e. PSA level, DRE and histology reports. For each core the mean PIRADS score of the core was used. Sensitivity, specificity and accuracy were assessed for different thresholds. Logistic regression analyses were performed between PIRADS scores and histopathology. A P ≤ 0.05 was considered to indicate statistical significance. We performed a core-wise receiver operating characteristic (ROC) analysis for each modality and for peripheral and transition zones separately, receiving the optimal threshold of PIRADS score for each modality. The positive predictive value (PPV) and NPV were calculated using a prevalence of prostate cancer of 36/100 000 men (Central Europe) according to the literature .
In 36 of 50 cases the initial MRI report was confirmed by biopsy. In the group of patients with histologically confirmed prostate cancer (29 men, Table 1), 22 patients (76%) were also described as ‘suspicious’ on mpMRI. In all, 26 men had intermediate- or high-risk disease on histopathology . At the time of biopsy the PIRADS score was not published, so our radiologists used a 3-point scale of ‘highly’, ‘questionable’ and ‘not suspicious’ findings. In the group of patients without prostate cancer on biopsy 14 patients (63%) were described as ‘not suspicious’ on mpMRI.
In all, 139 of 1177 cores (12%) were positive for prostate cancer. From these cores, 60% were localised in the peripheral zone of the prostate. The PIRADS score was significantly higher in cores positive for prostate cancer as compared with negative cores (Tables 3, 4). There was significant correlation between PIRADS scores andhistopathology in every modality but with limited clinical usefulness for MRS.
Table 3. Detailed evaluation of PIRADS score per core in each modality and pathology result
Table 4. Logistic regression analysis
Average PIRADS score (negative/positive histology)
Results of the ROC analysis for the peripheral zone for T2W, DWI, DCE and MRS (from top left to bottom right) are shown in Fig. 2, whereas results for the transition zone cores are shown in Fig. 3. Specificity for T2W with a PIRADS score threshold of 2 was 90% in the peripheral zone and 97% in the transition zone, respectively. For DWI with a PIRADS score threshold of 3, specificity was 98% in the peripheral zone and 97% in the transition zone. However, specificity decreased for DCE (80%/93%) and MRS (76%/83%). Sensitivity was low for all modalities (12–53%). The accuracy for each modality and details of the ROC analysis for different thresholds are listed in Table 5. Thresholds were rounded to the next integral value due to the discrete character of the PIRADS score.
Table 5. Results from ROC analysis for each modality and PIRADS score thresholds
AUC peripheral (Sens./Spec./PPV/NPV/accuracy)
AUC transition (Sens./Spec./PPV/NPV/accuracy)
AUC, area under the ROC curve.
Threshold = 2
Threshold = 4
Threshold = 5
Threshold = 3
Threshold = 4
Threshold = 5
Threshold = 2
Threshold = 4
Threshold = 5
Threshold = 2
Threshold = 4
Threshold = 5
Recently, mpMRI has been established as a diagnostic tool in the management of prostate cancer [14-16]. The potential of MRI/TRUS image fusion may improve biopsy guidance in the near future and may replace ‘blind’ systematic TRUS-guided biopsy protocols if validated in multicentre trials . Presently, five MRI/ultrasound fusion biopsy platforms are approved by the USA Food and Drug Administration (FDA) . Recently, Marks et al.  published their results derived from 360 patients, who were examined with one of these fusion systems (Artemis) and concluded that such fusion devices provide an accurate and efficient way to diagnose and manage prostate cancer in an outpatient setting. Vargas et al.  published a study on MRI predicting biopsy findings in a cohort of patients under active surveillance. The authors found that an overall PIRADS score of 5 was highly sensitive for upgrading of Gleason score on confirmation biopsy (87–98%). The patient cohort in the present study was less homogeneous with 45% of patients undergoing repeat prostate biopsy and only three patients under active surveillance. Nevertheless, we found a significant correlation between PIRADS scores and pathology (P < 0.05) for MRI modality. Furthermore, MRI reports and histopathology matched in 72% of cases. This renders mpMRI a valuable tool in the assessment of patients with known or suspected prostate carcinoma. However, the PPVs of the PIRADS scores, as described in Table 5, are very low. Thus, further efforts are required to increase the diagnostic accuracy and especially the sensitivity of mpMRI.
A major challenge in prostate cancer diagnostics is how to approach the problem of over-diagnosis of clinically insignificant disease [18-20]. A reduction would have a substantial impact from a socioeconomic point of view and would reduce psychological distress in patients with low-risk disease. In the present cohort, 26 of 29 patients were diagnosed with high- or intermediate-risk disease (NCCN classification ). Thus, mpMRI and standardised reporting via PIRADS score seems to provide useful information to the urologist in targeting the most ‘significant’ regions. Han et al.  recently published results, which showed that TRUS-guided biopsies failed to reliably target prior determined areas of interest. The rate of nearly 90% high- or intermediate-risk disease in the present cohort shows that MRI-targeted TRUS-guided biopsies may be the appropriate tool to transfer the benefit of standardised PIRADS-based reporting into urological practice by targeting areas with high PIRADS scores.
Rosenkrantz and Taneja  and Stoianovici  described the advances in imaging and targeted biopsy as a necessary precondition for individualised prostate cancer treatment. Standardised PIRADS-based reporting may allow urologists using MRI/TRUS fusion systems to target the most aggressive lesion (dominant intraprostatic lesion) for biopsy and further treatment planning. Standardised reports are a useful tool for patient follow-up under active surveillance. Nowadays ‘active surveillance’ protocols consist of PSA measurements at 3-month intervals and repeat biopsies . However, a significant number of upgrading is missed using these approaches [20, 23]. The use of mpMRI and standardised reporting via PIRADS scores may provide more detailed information to enhance prognostic nomograms and to increase patient safety. The number of patients under active surveillance in the present study is very low, but looking at the ROC analyses in the present study, a threshold of 2 or 3 in PIRADS scores per modality showed excellent specificity for T2W and DWI images regardless of the prostate region observed. All modalities had NPVs of 99% when a PIRADS score threshold of 2 (for T2W, DCE, and MRS) or 3 (for DWI) was used. Considering these results in light of the controversial discussion about the utility of PSA screening, mpMRI may become a valuable tool in pre-biopsy diagnostic evaluation, if validated in multicentre trials. For active surveillance , we agree that results from mpMRI must be further validated by, ideally transperineal, prostate biopsies to omit re-biopsies in active surveillance protocols.
Umbehr et al.  published a review on the combined use of MRI and MRS and reported higher sensitivities and specificities than in the present study. For MRS our inferior results may be explained by differences in the use of endorectal coils. Our standard protocol at 3.0 T does not include an endorectal coil for improved patient comfort; however, this results in increased movement of the prostate during MRS with concomitant increased motion artefacts and reduced signal-to-noise ratio. Vargas et al.  published their results using DWI for detection of prostate cancer. In the peripheral zone, they reported similar areas under the ROC curves (0.76–0.79). Analysis of the transition zone was not performed due to few prostate cancers in this area in their cohort.
For DCE, Sciarra et al.  found significant differences in reported sensitivities and specificities, due to a lack of standardised acquisition protocols and analytic models. Therefore, the optimal DCE protocol is still debatable today. In the present study, we did not calculate detailed DCE parameters, i.e. the vascular transfer constant (ktrans), but assigned a PIRADS score according to visualised contrast kinetic curves (ESUR guidelines). The interpretation of DCE with the help of PIRADS scoring showed imprecise results in the present data with sensitivity and specificity of 53% and 80%, respectively. This clearly indicates further improvement is needed. Additional technologies like ‘computer-aided diagnostics’ may allow more precise interpretation of imaging data. Our institution and many others are already working on software solutions in cooperation with computer scientists in this field.
In addition to the retrospective design of the present study, one drawback might be deformation of the prostate when comparing MRI data with intraprocedurally generated 3D-biopsy data. mpMRI were performed without an endorectal coil but during MRI/TRUS fusion biopsy the TRUS probe in the rectum may contort the prostate. This possible deviation may bias our core-by-core analysis. However, a previously performed evaluation of this ‘procedural targeting error’ in ex vivo phantoms showed only a minor deviation below 1 mm . The exact percentage of infiltration of prostate cancer per single core was not reported in the present study. More detailed histology reporting with exact percentages of infiltration per core would enable more precise statistical analyses.
In conclusion, the present results show that the PIRADS score is feasible for routine clinical use and allows standardised reporting to the referring urologist. It can be used as a decision support system for targeting of suspicious lesions. mpMRI has a high NPV for detection of prostate cancer and might be a valuable tool in pre-biopsy diagnostic evaluation.
We thank the European Foundation for Urology (EFU) for supporting this study.