5059
Optimal temporal resolution for accurate AIF measurement and error-constrained pharmacokinetic modelling of DCE data1National Centre for Advanced Medical Imaging (CAMI), St James Hospital / School of Medicine, Trinity College University of Dublin, Dublin 8, Ireland, 2Department of Radiology, Mayo Clinic, Rochester, MN, USA, Rochester, MN, United States
Synopsis
A highly-controlled and validated phantom-based method was used to investigate the effects of acquisition temporal resolution (Tres) on the arterial input function (AIF) measurement accuracy and precision for DCE-MRI. The propagation of these AIF measurement errors into errors in pharmacokinetic modelling parameter values could thus also be investigated. Guideline Tres values which can be used to constrain errors in Ktrans,
Introduction
The arterial input function (AIF) is a measure of change in contrast agent (CA) concentration over time in a blood vessel feeding the tissue of interest. Good characterization of the AIF is critical for accurate, patient-specific pharmacokinetic (PK) modelling using e.g. the Tofts model [1]. Despite the fact that the temporal resolution (Tres) at which the data are acquired is known to have a strong effect on PK parameter output accuracy, there is currently a lack of consensus as to the optimal Tres to use with DCE-MRI, with widely-varying values reported in the literature [typically 2-30s] [2]. In the present work a fully-validated anthropomorphic DCE-MRI phantom device was utilized, wherein precisely known and controlled time-varying CA concentration-time curves (CTCs), mimicking those observed in patient data, could be repeatedly produced and presented to the scanner for measurement [3]. The aim of the study was to comprehensively investigate the effects of Tres on AIF measurement and PK parameter output accuracy and precision.Methods
A physiologically-relevant AIF curve-shape was established from in-vivo patient data acquired at 3T (Achieva, Philips, Netherlands) using an 8-channel detector coil and an endorectal coil (Tres=3.1s, 3D-SPGR, TR/TE=5.5/2.0ms, α=15°, FOV=256x256x60mm3, voxels=1x1x6mm3, SENSE=2). The ground truth AIF, along with a ground truth tissue-mimicking CA CTC, were established for the phantom system using a previously-described purpose-built optical imaging system [3].
The ground truth AIF (shown in Figure 1), together with a second, tissue-mimicking CTC, were subsequently repeatedly-measured (x10) across 5 days using a 32-channel detector coil across the Tres range [1.2-30.6s], providing 250 datasets for analysis (3D-SPGR, TR/TE=3.5/1.6ms, α=23°, FOV=224x224x18mm3, voxels=1x1x6mm3, NSA=[1-25]). B1+ maps were also acquired using a dual-steady-state sequence (FOV=224x224x18mm3, voxels=1x1x6mm3, TR1/TR2/TE = 30ms/150ms/2ms, α=60°) and voxel-wise flip angle correction performed [4].
A number of AIF curve-shape features were quantified from each dataset, namely: time-to-peak 1 (TTP1), time-to-peak 2 (TTP2), CA concentration at peak 1 (P1) and 2 (P2), and the Wash-in Slope (0-TTP1) and Wash-out Slope (60-360s), as illustrated in Figure 1. PK modelling was then performed using a linear implementation of the standard Tofts model [5], with PK parameter errors calculated against precisely-known ground truth values.
Results
Figure 2 shows a temporally-aligned enlarged section of the AIF curves measured with Tres = [1.2-30.6s] at t = [-20-60s] relative to the ground truth AIF. Errors in the derived AIF curve-shape features are plotted in Figure 3. All errors in AIF curve-shape feature measurements were minimized using the fastest protocol tested (Tres=1.2s; all errors ≤9%), with most exhibiting relatively consistent errors up to Tres = 5s, increasing thereafter (particularly strongly for TPP1). The propagation of these errors into the resulting PK parameter values can be seen in Figure 2, where again errors remain fairly consistent up to a Tres of approximately 5s, above which Ktrans and kep increase strongly. Errors in ve remained relatively constant across the Tres range, as expected.
The data in Figure 5 shows the Tres values which were required to constrain errors in the measurement accuracy of Ktrans, kep and ve within arbitrary limits: ≤5%, ≤10%, ≤15%, and ≤20% were chosen to illustrate these trends.
Discussion
The results presented herein demonstrate the dramatic effect that an inappropriate acquisition Tres has on the measurement of the AIF, and by extension on the PK parameter values derived from the data. Quantification of these errors was made possible by a priori knowledge of the ground truth values and provides a framework upon which clinical protocols can be built: for example, while the ground truth can never be known in vivo, it is clear that using a Tres value > 10s with this DCE acquisition sequence will result in significant errors in PK parameter values. Knowledge of the ground truth values allowed for absolute quantification of these errors. Careful consideration should be made when selecting an appropriate Tres for a particular DCE application, which should be informed by the physiology of interest (i.e. the type of time-varying signal to be measured). In this work, Tres values suitable for a typical abdominal/pelvic DCE examination where measurement of the patient AIF is desired (i.e. if PK modelling is to be performed) whereby PK modelling errors can be constrained within pre-determined limits.Conclusion
This study comprehensively and quantitatively investigated the relationship between the fidelity of AIF curve-shape-feature measurements and the acquisition Tres, providing insight into how errors in AIF measurement propagate to errors in PK parameter outputs. Minimum Tres values are reported within which errors can be constrained below pre-determined values for this specific DCE protocol; these results can inform both scientific and clinical practice.Acknowledgements
Supported by Irish Cancer Society Research Scholarship [CRS13KNI]References
[1] Tofts PS, Brix G, Buckley DL, Evelhoch JL, Henderson E, Knopp MV, et al. Estimating kinetic parameters from dynamic contrast-enhanced T(1)-weighted MRI of a diffusable tracer: standardized quantities and symbols. J Magn Reson Imaging. 1999;10:223-32.
[2] Rosenkrantz AB, Geppert C, Grimm R, Block TK, Glielmi C, Feng L, et al. Dynamic contrast-enhanced MRI of the prostate with high spatiotemporal resolution using compressed sensing, parallel imaging, and continuous golden-angle radial sampling: Preliminary experience. J Magn Reson Imaging. 2015;41:1365-73.
[3] Knight SP, Browne JE, Meaney JF, Smith DS, Fagan AJ. A novel anthropomorphic flow phantom for the quantitative evaluation of prostate DCE-MRI acquisition techniques. Phys Med Biol. 2016;61:7466-83.
[4] Yarnykh VL. Actual flip-angle imaging in the pulsed steady state: a method for rapid three-dimensional mapping of the transmitted radiofrequency field. Magn Reson Med. 2007;57:192-200.
[5] Murase K. Efficient method for calculating kinetic parameters using T1-weighted dynamic contrast-enhanced magnetic resonance imaging. Magn Reson Med. 2004;51:858-62.
Figures
