Synopsis

Keywords: Data Analysis, DSC & DCE Perfusion

Recent advance in data acquisition makes fast DCE-MRI data acquisition feasible. This work investigated the impact of DCE-MRI temporal resolution on the data’s pharmacokinetic modeling and the water exchange effect imparted into the model parameters. Using both the Tofts model and the water-exchange sensitized Shutter-Speed model, our results showed that DCE data with higher temporal resolution (shorter intersample interval) is beneficial in model parameter precision and in quantifying transcytolemmal water exchange effect. The later may offer additional lesion-detection specificity in clinical prostate MRI.

Introduction

Pharmacokinetic modeling of prostate dynamic contrast-enhanced MRI (DCE-MRI) often employs the Tofts model.¹ The well-known finite transcytolemmal water exchange effect on prostate DCE-MRI modeling has been investigated with limited success.^2-4 The goal of this work is to investigate the impact of DCE-MRI temporal resolution on the data’s pharmacokinetic modeling and the water exchange effect imparted therein. Using both the Tofts model¹ and the Shutter-Speed model^3,4, our results show that DCE data with higher temporal resolution (shorter intersample interval) is beneficial in model parameter precision and in quantifying transcytolemmal water exchange effect. The later may offer additional lesion-detection specificity in clinical prostate MRI.

Methods

As part of two funded research projects, clinical multiparametric MRI (mpMRI) data with whole mount pathology were retrospectively analyzed after local IRB approvals. All MRI data were acquired at 3T scanners with endorectal RF coils. A standard dose of 0.1 mmol/kg gadolinium-based contrast agent (CA) was used in all DCE scans. Nineteen “slow” DCE data sets (intersample interval 8.51-12.45 s, with 12/19 at 8.67 s) from a Philips 3T scanner and 19 “fast” DCE data sets (intersample interval 4.95 - 6.33 s, with 14/19 at 4.95 s) from a Siemens 3T were included in this data analysis. Lesion and normal appearing regions of interest (ROIs) were drawn on post-CA DCE images with referencing to T2-weighted images and apparent diffusion coefficient (ADC) maps. Each DCE ROI time-course data was modeled with both the fast-exchange-limit (FXL) Tofts model¹ and the fast-exchange-regime (FXR) Shutter-Speed Model (SSM).^3,4 To facilitate the quantitative evaluation of temporal-resolution on prostate DCE modeling and the effect of transcytolemmal water exchange on model parameters, differences between lesion and normal appearing tissues of the parameters are calculated. For example, the FXL K^trans difference between lesion (les) and normal-appearing (NA) tissue is simply: K^trans (les) – K^trans (NA). In addition, following previous published method⁴, the K^trans difference between the two models { ΔK^trans = K^trans (FXR) - K^trans (FXL) } was calculated as an indirect approach in quantifying water exchange effect. Then, the difference of this parameter, { ΔK^trans (les) - ΔK^trans (NA)}, is summarized. Thus, the ΔK^trans approach has two differences built-in, that for K^trans difference between the models, and the 2^nd difference between lesion and NA of the derived parameter.

Results

All MRI visible lesions were confirmed by the whole mount pathology slides. Figure 1 shows mpMR images and a digitized whole mount pathology slide: a, axial T2-weighted image; b, ADC map; c, post-CA DCE image with color ROIs for lesion (blue) and NA (green). The orange arrows in a) – c) indicate the MRI suspicious lesion. The digital pathology slide (d) is roughly at the same level as that of the MRI slice and the general agreement of the Gleason grade 5 (GG5) lesion to that seen on MRIs confirms the imaging findings. Figure 2 compares the Tofts model’s K^trans difference in prostate lesion discrimination for the two groups of DCE-MRI data: slow (long intersample interval) and fast (short intersample interval) data acquisition (ACQ). The K^trans difference between the lesion and NA ROIs { = K^trans (les) - K^trans (NA) } for each subject is calculated and shown (circle for slow ACQ; diamond for fast ACQ). Apart from subject cohort heterogeneity, the slow ACQ showed a much larger range of K^trans difference than that of fast ACQ. Figure 3 shows the box and whisker plot similar to that of Fig. 2, but for the ΔK^trans difference between lesion and NA, { = ΔK^trans (les) - ΔK^trans (NA)}. Again, apart from the subject cohort difference, more ΔK^trans difference data points for slow ACQ are closer to zero with a few being negative. Using a ΔK^trans of 0.03 min^-1 as an example for cutoff, 6 out of 19 data points fell below the threshold for the slow ACQ group, while the number for fast ACQ is 4 out of 19. In addition, there is no negative ΔK^trans difference for the fast ACQ group.

Discussion

In this work, the implication of DCE-MRI temporal resolution on model parameters are investigated. Apart from subject cohort difference, results here show that fast ACQ inherently improves the temporal alignment of the time-courses between the arterial input function and the tissue. This results in greatly reduced variance in K^trans difference (Fig.2). It is generally observed that cancerous tissues are leakier than the normal-appearing counterpart. This implies more prominent water exchange effect is expected to impart into lesion DCE-MRI. Slow ACQ will also diminish this effect expressed in ΔK^trans difference (Fig. 3). Manifested in ΔK^trans for more appreciable water exchange effect, a greater number of ΔK^trans difference data points were above a threshold and no negative value resulted from the fast ACQ group. These mostly likely imply fast ACQ reduces K^trans variance and also results in a more realistic reflection of increased water exchange effect seen in malignant tissue. Findings from this work are expected to increase the lesion detection specificity in DCE-MRI. Combining this and more readily available fast DCE sequences like the golden-angle radial sparse parallel MRI⁵ in clinical settings, the relevance of DCE-MRI in prostate mpMRI is expected to improve at fast ACQ. Generally reduced signal-to-noise ratio has to be considered when designing fast ACQ.