Effect of temporal resolution on diagnostic performance of DCE-MRI of the prostate
Ahmed Othman1, Florian Falkner1, Jakob Weiss1, Stephan Kruck2, Robert Grimm3, Petros Martirosian1, Konstantin Nikolaou1, and Mike Notohamiprodjo1

1Department of Diagnostic and Interventional Radiology, University Hospital Tübingen, Tübingen, Germany, 2Department of Urology, University Hospital Tübingen, Tübingen, Germany, 3Siemens Healthcare, Siemens Healthcare, Erlangen, Germany


In DCE-MRI of the prostate, a high temporal resolution (TR) is recommended. Nonetheless, no studies on the effects of TR on diagnostic accuracy of DCE-MRI of the prostate are available. In this study we examined the effect of TR of DCE-MRI on its diagnostic accuracy for detection of potentially malignant lesions. Our results indicate that TR>10s / timepoint leads to spurious perfusion estimates and therefore deteriorates the diagnostic accuracy for identification of potentially malignant prostate lesions.


To assess the effect of temporal resolution on semi-quantitative and pharmacokinetic parameters from dynamic contrast-enhanced MRI (DCE-MRI) and their diagnostic accuracy regarding the detection of potentially malignant prostate lesions.

Material and methods

60 consecutive male patients (age, 64.5 ± 7.0 years) with clinically suspected prostate cancer were included. All patients underwent multi-parametric MRI of the prostate (T2-w, DWI-w and DCE-MRI) on a 3T MRI scanner. Patients were divided into 2 groups depending on PI-RADS score of the detected lesions (group A: PI-RADS ≤3, n=30; group B: PI-RADS >3, n=30). For DCE-MRI, a CAIPIRINHA-Dixon-TWIST Volume-Interpolated Breath-Hold Examination (CDT-VIBE) sequence was acquired (1). The parameterization of this spoiled gradient echo sequence included a bipolar dual-echo readout with echo times TE1 = 1.51 ms and TE2 = 2.62 ms for water-fat separation using the Dixon method. Therefore, a repetition time of TR = 4.6 ms, an FA of 9°, and a spatial resolution of 3 mm x 1.2 mm x 1.2 mm for a 256 x 144 matrix were chosen. A parallel-acquisition-technique (PAT-) undersampling factor of 2 was applied. Furthermore, TWIST view-sharing was applied, in order to increase temporal resolution between subsequent repetitions. The temporal resolution resulted in 5 s/timepoint. After pre-contrast dual-flip-angle T1 mapping (FA= 2°, 15°), DCE-MRI was acquired in 4:10 minutes (50 phases of 5 s) with body-weight-adapted administration of contrast agent (Gadobutrol, Bayer Healthcare, Berlin, Germany). 6 DCE-MRI series with different temporal resolutions ranging from 5 s to 30 s per timepoint were retrospectively generated from the original datasets (Figs. 1 and 2). Semi-quantitative parameters (i.e. Wash-In, Wash-Out and TTP) as well as pharmacokinetic parameters (i.e. Ktrans, Kep and ve) were calculated for the different temporal resolutions. Student’s t-tests were performed to assess differences of pharmacokinetic parameters between both groups. Repeated-measures (rm-) ANOVA were conducted to compare semi-quantitative and pharmacokinetic parameters from the 6 different DCE-MRI series. Receiver operating characteristic curve (ROC) analysis with areas under the curve (AUC) and the corresponding 95% confidence intervals (95%-CI) were used for assessment of diagnostic accuracy of the different series for the differentiation between both lesion groups.


A significant effect of temporal resolution was found on Wash-In (p<.001). Series with temporal resolution lower than 10 s/timepoint showed significantly lower Wash-In values with more pronounced effects in group B compared to group A (Fig. 3). For 30s series, the differences between both groups diminished reaching insignificant levels (p=.052), resulting in a significant decrease of the diagnostic accuracy of Wash-In (AUC: .609, 95%-CI .451 – .766;p<.015, Fig. 4). No significant effects were detected on Wash-Out. For TTP, a significant effect of temporal resolution was detected (p<.001) with significantly increasing TTP levels for all down-sampled series compared to the original 5 s series. These effects did not impact the diagnostic accuracy of TTP. No significant effects of temporal resolution were detected on pharmacokinetic parameters (p<.112).


In DCE-MRI of the prostate, temporal resolution affects the diagnostic performance of semi-quantitative parameters. For a sufficient detection of malignant prostate lesions on DCE-MRI, a temporal resolution of at least 10 s/timepoint or higher is recommended.




1. Othman AE, Martirosian P, Schraml C, et al. Feasibility of CAIPIRINHA-Dixon-TWIST-VIBE for dynamic contrast-enhanced MRI of the prostate. Eur J Radiol. 2015.


Figure 1: Flowchart of data acquisition, retrospective downsampling and parameter estimation.

Figure 2: Pathologically confirmed malignant lesion in the right peripheral zone with typical wash-in on DCE-MRI (middle row). This lesion was classified as PI-RADS 5. Perfusion enhancement curves of the lesion derived from datasets with temporal resolutions ranging between 5 s and 30s are given in the lower row.

Figure 3: Semi-quantitative and pharmacokinetic parameters for both groups from the different series. Please note the diminishing difference between Groups A and B at lower temporal resolutions.

Figure 4: ROC curves for diagnostic accuracy for discrimination between both lesion groups based on Wash-In from 5s and 30s series. Please note the lower diagnostic accuracy (AUC) of 30 s datasets compared to the 5s datasets.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)