E. H. J. Hamoen MD, Presenter: Nothing to Disclose

Les Thompson, Abstract Co-Author: Nothing to Disclose

Fred Witjes MD, PhD, Abstract Co-Author: Nothing to Disclose

Maroeska M. Rovers PhD, Abstract Co-Author: Nothing to Disclose

Jelle O. Barentsz MD, PhD, Abstract Co-Author: Nothing to Disclose

To evaluate the accuracy and interobserver variability of the final PI-RADS classification based on a dominant MR-sequence compared to the often used single-modality sum score.

223 biopsy-naïve men suspected of having prostate cancer were included in a prospective clinical trial. All men underwent a 3T mp-MRI, including T2-weighted imaging (T2WI), diffusion-weighted imaging (DWI), and dynamic contrast-enhanced (DCE) MRI. Histology of all lesions was obtained by in-bore MR-guided biopsy followed by standard TRUSGB in MR-positive men, or only standard TRUSGB in MR-negative men. All MRI sequences were co-read independently by 2 investigators. Any discrepancies were resolved by consensus. Both investigators assigned single-modality scores and an overall “dominant” PI-RADS to all lesions, of which the latter was based on DWI in peripheral zone lesions, and on T2WI in transitional zone lesions. Single-modality sum-scores were calculated and compared to overall “dominant” PI-RADS. 2x2 contingency tables were created to calculate sensitivity, specificity, PPV and NPV. Proportions of agreement were calculated.

Best accuracy rates were reached using the overall “dominant” PI-RADS with a threshold of ≥ 4. Reader 1 and respectively 2 achieved a sensitivity of 89.8% (97/108) and 81.5% (88/108), specificity of 86.1% (99/115) and 86.1% (99/115), PPV of 85.8% (97/113) and 84.6% (88/104), and NPV of 90.0% (99/110) and 83.2% (99/119) for detecting significant prostate cancer. Using the sum score with a threshold of ≥ 10, reader 1 and respectively 2 achieved a sensitivity of 89.8% (97/108) and 81.5% (88/108), specificity of 73.9% (85/115) and 80.9% (93/115), PPV of 76.4% (97/127) and 80.0% (88/110), and NPV of 88.5% (85/96) and 82.3% (93/113) for detecting significant prostate cancer. Proportions exact agreement were 73.1% for overall “dominant” PI-RADS, 44.4% for DCE-MRI, 51.1% for T2WI, and 56.5% for DWI.

The overall “dominant” PI-RADS is a robust interpretation score for mp-MRI to detect significant cancer with good inter-reader agreement, which outperforms the commonly used single-modality sum-score.

Overall "dominant" PI-RADS accurately detects significant prostate cancer with good interreader agreement and is recommended in the evaluation of mp-MRI in men suspicious for prostate cancer instead of the single-modality sum score.

Hamoen, E, Thompson, L, Witjes, F, Rovers, M, Barentsz, J, Evaluation of PI-RADS for Multi-parametric Prostate MRI: How to Improve the Overall Score?.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14016933.html