Prostate cancer detection with multi-parametric MRI : PI-RADS version 1 versus version 2
Zhaoyan Feng1, Xiangde Min1, and Liang Wang1

1Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China, People's Republic of

Synopsis

The new PI-RADS version 2 classification (PI-RADS v2) was proposed together with the European Society of Urogenital Radiology (ESUR) and the American College of Radiology (ACR) in December 2014. In contrast to PI-RADS v1, the v2 regulate how to classify final PI-RADS score. for routine clinical use, test of the validity of v2, including its sensitivity and specificity for prostate cancer (PCa) detection should raise concerns, and literature of them less. So, the purpose of our study was to compare the diagnostic performance of v1 and v2 for the detection of PCa.

Purpose

To determine if PI-RADS v2 scoring system improve diagnostic accuracy of multi-parametric MRI (mp-MRI) of the prostate in comparison to PI-RADS v1.

Methods

This retrospective study was approved by the institutional review board. Four hundred and one consecutive patients clinically suspicious prostate cancer undergoing 3.0 T mp-MRI (T2W + DWI + DCE) before transrectal ultrasound (TRUS) guided systematic biopsies were included between June 2013 and July 2015. Two radiologists with 4 and 3 years of experience and a minimum of 1000 prostate MRI readings analyzed the imaging studies in consensus.For reporting the location of each lesion, the prostate gland was divided into sixteen regions of interest on MRI scans[1]. In a first step, all the lesions are rated on a score from 1 to 5 in each of the three MRI sequences (T2WI, DWI, and DCE-MRI) according to the PI-RADS v1. Analysis for the present study was by patient not prostate sector and the highest sum PI-RADS score lesion for each mpMRI scan should be identified. Then we obtained the overall PI-RADS score (1-5) by classifying the sum-score according to the algorithm proposed by Rothke et al[2] . In a second step, the readers again scored the lesions according to the PI-RADS v2 score and generate a PI-RADS score (1-5) . Each patient was classified as having TZ or PZ cancer if more than 70% of the cancer volume was present in a particular zone[11]. Visual examples for PI-RADS v1 and v2 were provided(Fig.1). After all MR imaging readings and subsequently TRUS-guided biopsy were completed, comparison analysis was performed between correlated foci on MRI and histologic findings.Receiver operating characteristics (ROC) were calculated for statistical analysis. Sensitivity, specificity, diagnostic accuracy was compared using McNemar test.

Results

The mean age of our study population was 64.39±8.98 (SD) years (range, 34-88 years). 150 patients were diagnosed as PCa, of which 95 were located in the peripheral zone and 55 in the transitional zone. When we pooled data from both zones, the area under the curve was 0.889 for PI-RADS v1 and 0.942 for v2 (P = 0.0001). For the PZ, the area under the curve was 0.916 for the v1 and 0.930 for the v2 (P = 0.3065). For the TZ, the area under the curve was 0.841 for the v1 and 0.971 for the v2 (P < 0.001) (Table 1). ROC analysis was performed to determine the cutoff value which optimized sensitivity, specificity, and accuracy for PCa detection (Table 2). With a cutoff value of 4 (scores of 4 and 5 were used as a positive mpMRI). When we considered the two zones combined, PI-RADS v2 demonstrated significantly higher sensitivity (96% [144 of 150]) vs 84.67% [127 of 150], P < 0.001), similar specificity (84.06% [211 of 251] vs 81.27% [204 of 251], P = 0.311), and significantly higher accuracy (88.53% [355 of 401] vs 82.54% [331 of 401], P = 0.002) in comparison with the v1. We also evaluated the performance of the PI-RADS scoring system in central gland lesions and in peripheral zone lesions separately. For the PZ, there was no difference in sensitivity, specificity and accuracy using PI-RADS v1 compared to PI-RADS v2 (all P > 0.05). For the TZ, there was a significant improvement in the sensitivity for PCa detection using PI-RADS v2 (96.36% [53 of 55]) compared to PI-RADS v1 (76.36% [42 of 55]), P = 0.003; there was no difference in the specificity using PI-RADS v2 (90.24% [74 of 82]) compared to PI-RADS v1 (84.15% [69 of 82]), P = 0.227; Moreover, overall accuracy improved using PI-RADS v2 (92.70% [127 of 137]) vs PI-RADS v1 (81.02%[111 of 137]), P = 0.002.

Discussion and Conclusion

In this study, summed PI-RADS v1 and v2 scores both showed good capabilities in cancer detection. Compared to diagnostic reporting with v1, it was able to be shown that the diagnostic accuracy increased from 0.82 to 0.88 with the use of v2 for all patients in our study cohort. The diagnostic accuracy value for the detection of TZ PCa was 0.81 for v1 and 0.92 for v2. The summed PI-RADS v2 outperformed the v1 in assessment of the TZ and the two zones, while exhibited similar sensitivity, specificity, and accuracy in assessment of the PZ. Our results indicate that the diagnostic performance depend on cancer location depiction. Our study was in accordance with other studies[3;4].

All the original v1 and v2 showed good diagnostic performance for the detection of prostate cancer. Although, in the TZ, performance was better with the v2 than the v1 scale.

Acknowledgements

References

[1] Dickinson, et al. Eur Urol 2011; 59:477-494.

[2] Rothke, et al. Rofo 2013;185:253-261.

[3] Baur, et al. Invest Radiol 2014; 49:411-420

[4] Ginsburg, et al. J Magn Reson Imaging 2015; 41:1383-1393

Figures

Fig 1. PI-RADS v1 scores: 4 for T2WI, 4 for DWI, 4 for DCE, summed score = 12, this corresponds to a PI-RADS v1 4. PI-RADS v2 the area on T2WI and DWI corresponds to score 5 (larger than 1.5 cm), and DCE with "+", v2 overall score 5.

Table 1 AUC values of single PI-RADS scores for T2WI, DWI, DCE and for summed PI-RADS score for cancer detection in PZ, TZ, and both PZ and TZ.

Table 2 Diagnostic performance of PI-RADS v1 and v2 for PZ, TZ, and both PZ and TZ.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
0726