Value of an online PI-RADS v2.1 score calculator for assessment of prostate MRI

Borna K Barth; Katharina Martini; Stephan M Skawran; Florian A Schmid; Niels J Rupp; Laura Zuber; Olivio F Donati

doi:10.1016/j.ejro.2021.100332

Value of an online PI-RADS v2.1 score calculator for assessment of prostate MRI

Eur J Radiol Open. 2021 Feb 27:8:100332. doi: 10.1016/j.ejro.2021.100332. eCollection 2021.

Authors

Borna K Barth¹, Katharina Martini¹, Stephan M Skawran¹, Florian A Schmid², Niels J Rupp³, Laura Zuber⁴, Olivio F Donati¹

Affiliations

¹ Institute of Diagnostic and Interventional Radiology, University Hospital Zurich and University of Zurich, Raemistrasse 100, CH-8091 Zurich, Switzerland.
² Department for Urology, University Hospital Zürich, Rämistrasse 100, CH-8091 Zurich, Switzerland.
³ Department of Pathology and Molecular Pathology, University Hospital Zurich and University of Zurich, Raemistrasse 100, CH-8091 Zurich, Switzerland.
⁴ University of Zurich, Raemistrasse 78, 8006 Zurich, Switzerland.

Abstract

Purpose: To evaluate the value of a browser-based PI-RADS Score Calculator (PCalc) compared to MRI reporting using the official PI-RADS v2.1 document (PDoc) for non-specialized radiologists in terms of reporting efficiency, interrater agreement and diagnostic accuracy for detection of clinically significant prostate cancer (PCa).

Methods: Between 09/2013 and 04/2015, 100 patients (median age, 64.8; range 47.5-78.2) who underwent prostate-MRI at a 3 T scanner and who received transperineal prostate mapping biopsy within <6 months were included in this retrospective study. Two non-specialized radiology residents (R1, R2) attributed a PI-RADS version 2.1 score for the most suspect (i. e. index) lesion (i) using the original PI-RADS v2.1 document only and after a 6-week interval (ii) using a browser-based PCalc. Reading time was measured. Reading time differences were assessed using Wilcoxon signed rank test. Intraclass-correlation Coefficient (ICC) was used to assess interrater agreement (IRA). Parameters of diagnostic accuracy and ROC curves were used for assessment of lesion-based diagnostic accuracy.

Results: Cumulative reading time was 32:55 (mm:ss) faster when using the PCalc, the difference being statistically significant for both readers (p < 0.05). The difference in IRA between the image sets (ICC 0.55 [0.40, 0.68]) and 0.75 [0.65, 0.82] for the image set with PDoc and PCalc, respectively) was not statistically significant. There was no statistically significant difference in lesion-based diagnostic accuracy (AUC 0.83 [0.74, 0.92] and 0.82 [95 %CI: 0.74, 0.91]) for images assessed with PDoc as compared to PCalc (AUC 0.82 [0.74, 0.91] and 0.74 [95 %CI: 0.64, 0.83]) for R1 and R2, respectively.

Conclusion: Non-specialized radiologists may increase reading speed in prostate MRI with the help of a browser-based PI-RADS Score Calculator compared to reporting using the official PI-RADS v2.1 document without impairing interreader agreement or lesion-based diagnostic accuracy for detection of clinically significant PCa.

Keywords: PI-RADS 2.1; Prostate MRI; Prostate cancer; Reporting software.