How Prostate Cancer Growth Patterns Impact Detection and Interreader Agreement on Multiparametric MRI

This is the authors' abstract. We don't add key points for this paper.

Rationale and Objectives Multiparametric MRI (mpMRI) substantially improves the detection of significant prostate carcinoma (PCa) compared to systematic biopsy. Nevertheless, mpMRI can overlook aggressive forms of PCa. Recent studies showed, that infiltrative growth (INF) has less restricted diffusion. This study aims to explore the impact of growth patterns on the detection of lesions. Materials and methods This retrospective study analyzed 52 patients who underwent radical prostatectomy, with preoperative mpMRI. For each patient, one dominant lesion was identified on one whole-mount prostatectomy section. Two pathologists (P1, P2) independently classified the growth pattern whether as expansive (EXP) being defined with at least three 5mm² regions of interest consisting entirely of carcinoma without benign glands or else as infiltrative (INF). Two radiologists (R1, R2) independently classified selected lesions according to PI-RADSv2.1. based on pathological localization. Apparent diffusion coefficient (ADC) values were measured in correlation with matched histopathology slides. Interreader-agreement was evaluated using weighted Cohen's Kappa. The relationship between PI-RADS scores and pathological diagnoses was analyzed using logistic regression. Results Pathologic lesion characterization regarding growth patterns achieved almost perfect agreement (κ = 0.88), so did PI-RADS classification of mpMRI (κ = 0.90). PI-RADS scores correlated significantly with EXP growth patterns. Average ADC values were lower for EXP lesions (0.83×10 −3 mm 2 /s, CI: 0.72–0.94×10 −3 mm 2 /s) compared to INF lesions (0.97×10 −3 mm 2 /s, CI: 0.86–1.07×10 −3 mm 2 /s; p = 0.08). On T2 images, 8 of 28 (29%) INF lesions and 1 of 24 (4%) EXP lesions were not visible. Conclusion PCa missed on mpMRI more frequently demonstrate INF growth patterns. Lesions with EXP growth patterns show lower ADC values and have higher PI-RADS scores.