"Diagnostic accuracy and longitudinal outcomes of PI-RADS 3 lesions: A multi-reader comparison between biparametric...

To compare the diagnostic accuracy and longitudinal outcomes of PI-RADS 3 lesions assessed with biparametric (bpMRI) versus multiparametric MRI (mpMRI), in a multi-reader setting.

This retrospective single-center study included 196 patients, of whom 49 presented with at least one PI-RADS 3 lesion. Three radiologists with varying levels of experience independently reviewed both biparametric (bpMRI) and multiparametric (mpMRI) datasets. Diagnostic performance, inter-reader agreement, lesion evolution over time, and detection rates were analyzed in relation to lesion location and MRI protocol. Receiver operating characteristic (ROC) curve analysis, Cohen's kappa and intraclass correlation coefficients (ICCs) were used for statistical evaluation.

Reader 1 showed the highest diagnostic accuracy (AUC = 0.767), followed by Reader 2 (AUC = 0.637) and Reader 3 (AUC = 0.504). There were no significant differences between bpMRI and mpMRI for any reader and intra-reader concordance was excellent for all readers (κ > 0.8). Detection was significantly lower in the peripheral zone across all readers. Inter-reader agreement was substantial between Readers 1 and 2 (ICC = 0.704), and poor between the other pairs (ICC < 0.4). During follow-up, 84-92 % of PI-RADS 3 lesions remained stable, 6-8 % decreased, and 8-16 % increased in size. Seven lesions were biopsied based on PSA progression or dimensional growth, all of which were confirmed as clinically significant cancer.

BpMRI provided comparable diagnostic performance to mpMRI for PI-RADS 3 lesions. Reader experience and lesion location significantly affected detection. Most lesions remained stable over time, supporting surveillance strategies in selected cases.