A Multi-Institutional Comparison of Dynamic Contrast-Enhanced Magnetic Resonance Imaging Parameter Calculations

Abstract

Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) provides quantitative metrics (e.g. K^trans, v_e) via pharmacokinetic models. We tested inter-algorithm variability in these quantitative metrics with 11 published DCE-MRI algorithms, all implementing Tofts-Kermode or extended Tofts pharmacokinetic models. Digital reference objects (DROs) with known K^trans and v_e values were used to assess performance at varying noise levels. Additionally, DCE-MRI data from 15 head and neck squamous cell carcinoma patients over 3 time-points during chemoradiotherapy were used to ascertain K^trans and v_e kinetic trends across algorithms. Algorithms performed well (less than 3% average error) when no noise was present in the DRO. With noise, 87% of K^trans and 84% of v_e algorithm-DRO combinations were generally in the correct order. Low Krippendorff's alpha values showed that algorithms could not consistently classify patients as above or below the median for a given algorithm at each time point or for differences in values between time points. A majority of the algorithms produced a significant Spearman correlation in v_e of the primary gross tumor volume with time. Algorithmic differences in K^trans and v_e values over time indicate limitations in combining/comparing data from distinct DCE-MRI model implementations. Careful cross-algorithm quality-assurance must be utilized as DCE-MRI results may not be interpretable using differing software.

Figure 1

Plots of algorithm performance in a DRO with no noise for (a) K^trans and (b) v_e. The simulated values are on the x-axis, and the measured values from each algorithm are on the y-axis. The 45° line represents 100% accuracy of the measured values. Each color represents a different algorithm, and each shape represents a different v_e column in (a) and a different K^trans row in (b).

Figure 2

Heat maps of the percentage error for K^trans (top left) and v_e (bottom left) by algorithm in the 28 DROs with noise. The percentage error is defined using the formula ([measured − simulated]/simulated *100). The left side of the heat map is grouped by the timing interval used for the DRO (6 or 10 s), the timing offset used for the DRO (0 or 3 s for the 6 s timing interval, 0 or 5 s for the 10 s timing interval), and the SNR (0.18–1.8). The inset (top right) shows the K^trans and SNR values for each block in the heat maps. The maximum percentage error is defined as 100%, and the minimum percentage error is set to −100%. Any errors greater than the maximum percentage are also mapped as 100% error in color. Each DRO is differentiated by its sampling interval, timing offset, and SNR as determined by the S₀ and sigma value used to create the DRO.

Figure 3

Percentages of (a) K^trans and (b) v_e values removed from patient images. The boxplots for each algorithm include the percentages removed for all patients and contours.

Figure 4

Illustration of differences in K^trans (min⁻¹) maps exported by different algorithms for one axial DCE-MRI slice.