To evaluate the variability induced by the imager in discriminating high-grade (Gleason≥7) prostate cancers (HGC) using dynamic contrast-enhanced MRI.

We retrospectively selected 3T MRIs with temporal resolution<10 seconds and comprising T1 mapping from a prospective radiologic–pathologic database of patients treated by prostatectomy. Ktrans, Kep, Ve and Vp were calculated for each lesion seen on MRI using the Weinmann arterial input function (AIF) and three patient-specific AIFs measured in the right and left iliac arteries in pixels in the center of the lumen (psAIF-ST) or manually selected by two independent readers (psAIF-R1 and psAIF-R2).

A total of 43 patients (mean age, 63.6±4.9 [SD]; range: 48–72 years) with 100 lesions on MRI (55 HGC) were selected. MRIs were performed on imager A (22 patients, 49 lesions) or B (21 patients, 51 lesions) from two different manufacturers. Using the Weinmann AIF, Kep (P=0.005), Ve (P=0.04) and Vp (P=0.01) significantly discriminated HCG. After adjusting on tissue classes, the imager significantly influenced the values of Kep (P=0.049) and Ve (P=0.007). Using patient-specific AIFs, Vp with psAIF-ST (P=0.008) and psAIF-R2 (P=0.04), and Kep with psAIF-R1 (P=0.03) significantly discriminated HGC. After adjusting on tissue classes, types of patient-specific AIF and side of measurement, the imager significantly influenced the values of Ktrans (P=0.0002), Ve (P=0.0072) and Vp (P=0.0003). For all AIFs, the diagnostic value of pharmacokinetic parameters remained unchanged after adjustment on the imager, with stable odds ratios.

The imager induced variability in the absolute values of pharmacokinetic parameters but did not change their diagnostic performance.