[Show abstract][Hide abstract] ABSTRACT: The purpose of this work was to assess the reproducibility of diffusion imaging, and in particular the apparent diffusion coefficient (ADC), intra-voxel incoherent motion (IVIM) parameters and diffusion tensor imaging (DTI) parameters, across multiple centres using clinically available protocols with limited harmonization between sequences. An ice-water phantom and nine healthy volunteers were scanned across fives centres on eight scanners (four Siemens 1.5T, four Philips 3T). The mean ADC, IVIM parameters (diffusion coefficient D and perfusion fraction f) and DTI parameters (mean diffusivity MD and fractional anisotropy FA), were measured in grey matter, white matter and specific brain sub-regions. A mixed effect model was used to measure the intra- and inter-scanner coefficient of variation (CV) for each of the five parameters. ADC, D, MD and FA had a good intra- and inter-scanner reproducibility in both grey and white matter, with a CV ranging between 1% and 7.4%; mean 2.6%. Other brain regions also showed high levels of reproducibility except for small structures such as the choroid plexus. The IVIM parameter f had a higher intra-scanner CV of 8.4% and inter-scanner CV of 24.8%. No major difference in the inter-scanner CV for ADC, D, MD and FA was observed when analysing the 1.5T and 3T scanners separately. ADC, D, MD and FA all showed good intra-scanner reproducibility, with the inter-scanner reproducibility being comparable or faring slightly worse, suggesting that using data from multiple scanners does not have an adverse effect compared with using data from the same scanner. The IVIM parameter f had a poorer inter-scanner CV when scanners of different field strengths were combined, and the parameter was also affected by the scan acquisition resolution. This study shows that the majority of diffusion MRI derived parameters are robust across 1.5T and 3T scanners and suitable for use in multi-centre clinical studies and trials.
Full-text · Article · Apr 2015 · NMR in Biomedicine
[Show abstract][Hide abstract] ABSTRACT: The objectives are to examine the reproducibility of functional MR imaging in children with solid tumours using quantitative parameters derived from diffusion-weighted (DW-) and dynamic contrast enhanced (DCE-) MRI.
Patients under 16-years-of age with confirmed diagnosis of solid tumours (n = 17) underwent free-breathing DW-MRI and DCE-MRI on a 1.5 T system, repeated 24 hours later. DW-MRI (6 b-values, 0-1000 sec/mm(2)) enabled monoexponential apparent diffusion coefficient estimation using all (ADC0-1000) and only ≥100 sec/mm(2) (ADC100-1000) b-values. DCE-MRI was used to derive the transfer constant (K(trans)), the efflux constant (kep), the extracellular extravascular volume (ve), and the plasma fraction (vp), using a study cohort arterial input function (AIF) and the extended Tofts model. Initial area under the gadolinium enhancement curve and pre-contrast T1 were also calculated. Percentage coefficients of variation (CV) of all parameters were calculated.
The most reproducible cohort parameters were ADC100-1000 (CV = 3.26 %), pre-contrast T1 (CV = 6.21 %), and K(trans) (CV = 15.23 %). The ADC100-1000 was more reproducible than ADC0-1000, especially extracranially (CV = 2.40 % vs. 2.78 %). The AIF (n = 9) derived from this paediatric population exhibited sharper and earlier first-pass and recirculation peaks compared with the literature's adult population average.
Free-breathing functional imaging protocols including DW-MRI and DCE-MRI are well-tolerated in children aged 6 - 15 with good to moderate measurement reproducibility.
• Diffusion MRI protocol is feasible and well-tolerated in a paediatric oncology population. • DCE-MRI for pharmacokinetic evaluation is feasible and well tolerated in a paediatric oncology population. • Paediatric arterial input function (AIF) shows systematic differences from the adult population-average AIF. • Variation of quantitative parameters from paired functional MRI measurements were within 20 %.
Full-text · Article · Mar 2015 · European Radiology
[Show abstract][Hide abstract] ABSTRACT: To evaluate dynamic contrast-enhanced (DCE) magnetic resonance (MR) imaging for monitoring and assessing treatment response in patients with neuroendocrine liver metastases treated using yttrium 90 ((90)Y)-labeled octreotide ((90)Y-DOTATOC).
The study was approved by the local research and ethics committee and patient informed consent was obtained. Twenty patients with liver metastases from neuroendocrine tumors underwent T1-weighted DCE MR imaging of the liver before and at 2 months after intravenous (90)Y-DOTATOC treatment. Regions of interest were drawn around target lesions, as well as along liver outlines for each patient. A dual-input single-compartment model was used to compute parameters including fractional distribution volume and the arterial flow fraction. Pre- and posttreatment values were compared using Wilcoxon signed rank test. Treatment response was defined as showing a greater than 50% reduction in the nadir chromogranin A level within the 1st year after treatment. Pretreatment values of responders and nonresponders were compared using the Mann-Whitney test. A two-tailed P value of .008 or less, which accounts for multiple testing, was considered to indicate a significant difference.
In responders, tumor and whole liver distribution volume significantly increased after treatment (median tumor distribution volume, 0.182 vs 0.244; median whole liver distribution volume, 0.175 vs 0.207; P = .008). The pretreatment whole liver distribution volume was significantly lower in responders (median, 0.175 vs 0.248; P = .003), while pretreatment tumor arterial flow fraction was significantly higher in responders (median, 1.000 vs 0.7 ± 1, P = .006).
DCE MR imaging may be used to monitor the effects of peptide receptor radiolabeled targeted therapy in patients with neuroendocrine tumors liver metastases; a lower pretreatment distribution volume and high arterial flow fraction was associated with a better response to treatment.
[Show abstract][Hide abstract] ABSTRACT: Neuroendocrine hepatic metastases exhibit various contrast uptake enhancement patterns in dynamic contrast-enhanced MRI. Using a dual-input two-compartment distributed parameter model, we analyzed the dynamic contrast-enhanced MRI datasets of seven patient study cases with the aim to relate the tumor contrast uptake patterns to parameters of tumor microvasculature. Simulation studies were also performed to provide further insights into the effects of individual microcirculatory parameter on the tumor concentration-time curves. Although the tumor contrast uptake patterns can be influenced by many parameters, initial results indicate that hepatic blood flow and the ratio of fractional vascular volume to fractional interstitial volume may potentially distinguish between the patterns of neuroendocrine hepatic metastases.
No preview · Article · Jan 2011 · Magnetic Resonance in Medicine
[Show abstract][Hide abstract] ABSTRACT: Dynamic contrast enhanced magnetic resonance imaging (DCE-MRI) is a valuable tool for assessing treatment response to novel cancer therapeutics. With appropriate data acquisition, quantitative functional parameter estimates can be obtained by fitting a model to the data. This research focuses on applying a dual-input single-compartment pharmacokinetic model to breath-hold DCE-MRI imaging of the liver. In this paper, the use of two breath-holds, providing greater temporal information, is compared with a single breath-hold approach. Computer simulations are used to assess the accuracy, precision and sensitivity to input function errors obtained for parameters estimated from the two imaging protocols. Data from ten patients were analysed to assess the noise statistics obtained from the two breath-hold protocols. The noise statistics were used with a pharmacokinetic liver model to simulate data, from which the estimation accuracy, precision and sensitivity for the two protocols were assessed. Data from the ten patients were also analysed, and the estimates were compared with literature values. This work demonstrates the feasibility of obtaining functional liver perfusion estimates over a 3D volume using a sequential breath-hold protocol. The simulation results show that the protocol consisting of two images per breath-hold is to be preferred as it requires identical patient co-operation, but provides parameter estimates that have superior accuracy and precision.
No preview · Article · May 2009 · Physics in Medicine and Biology
[Show abstract][Hide abstract] ABSTRACT: Hepatic metastases are arterially supplied, resulting in an elevated hepatic perfusion index (HPI). The purpose of this study was to use dynamic contrast-enhanced (DCE) MR imaging to quantify the HPI of metastases and the liver before and after treatment with a novel antiangiogenic drug. Ten patients with known metastatic liver disease underwent DCE-MR studies. HPIs of metastases and whole liver were derived using regions of interest (ROIs) and calculated on a pixel-by-pixel basis from quantified changes in gadopentetate dimeglumine (Gd-DTPA) concentration. The HPI measurement error prior to treatment was derived by the Bland-Altman analysis. The median HPI before and after treatment with antiangiogenic drug BIBF 1120 were compared using the Wilcoxon signed rank test. Prior to treatment, the median HPI of metastases, 0.75 +/- 0.14, was significantly higher than that of the whole liver, 0.66 +/- 0.16 (p < 0.01). Bland-Altman reproducibility coefficients of the median HPI from metastases and whole liver were 13.0 and 5.1% respectively. The median HPI of metastases decreased significantly at 28 days after treatment with BIBF 1120 (p < 0.05). This pilot study demonstrates that HPI determined using quantified Gd-DTPA concentration is reproducible and may be useful for monitoring antiangiogenic treatment response of hepatic metastases.
No preview · Article · Jul 2008 · European Radiology
[Show abstract][Hide abstract] ABSTRACT: Introduction Fitting a dual-input kinetic model to DCE-MRI data from liver is necessary to accurately capture the concentration time-course (CTC) behaviour seen in this organ. A parameter is therefore required to describe the relative arterial and portal contributions of the dual-input contrast delivery to each voxel, and typically the onset time of contrast enhancement is also estimated for each voxel. In practice, estimates of the arterial-portal fraction parameter and the onset time are largely influenced by only a few data points close to the onset time, so errors in these parameters are typically highly correlated. Therefore, reducing errors in the onset time estimates will tend to reduce errors in the arterial-portal fraction parameter. The variability in the onset time estimates can be reduced by modelling and estimating the onset time as a single global parameter derived from all voxels simultaneously. Assuming this reduces the error in the onset estimate, it will also reduce the error in the arterial-portal fraction parameter estimates. The key question is therefore whether the differences between the true onset times and the global onset estimate are larger or smaller than the errors on voxel-wise onset time estimates. Since ground-truth is not available for in-vivo data this question cannot be answered directly, and so must be answered using statistical methods. Theoretically there will be variations in the enhancement onset time over the liver, and a priori this is likely to be more pronounced in diseased states. However, from a statistical viewpoint, it is the relationship between the data sampling rate and the onset time variation that determines the need to model the onset time as a local voxel-wise parameter or as a single global parameter for the whole organ, or region of interest. In this abstract we propose using various statistical information criteria to determine whether the onset time should be modelled as a global or local parameter for DCE-MRI liver data acquired using two specific time-sampling protocols. Methods Data Acquisition Protocol DCE-MRI data were acquired coronally on a 1.5T Siemens Avanto using a 3D FFE sequence under sequential breath-hold at expiration, which gives highly reproducible registration of the liver 1 . For protocol A 1 , each breath-hold image was acquired in 5.6 sec, followed by a 6.4 sec breathing gap, and 20 images were acquired giving a total time of 4 minutes. For protocol A 2 , two image volumes were acquired in 6 sec followed by a 6 sec breathing gap, and 40 images were acquired during the study. The imaging parameters for A 1 were TR/TE = 4.36/1.32 ms, FA = 24 o , 20 slices @ 5mm thick, and for A 2 were TR/TE = 3.28/1.10 ms, FA = 18 o , 12 slices @ 5mm thick, and for both NSA = 1, IPAT = 2, FOV = 350mm, 128×128 interpolated to 256×256 matrix. The dynamic scan was preceded by a calibration scan with the same parameters except FA = 2 o to enable the dynamic sequence to be converted to contrast agent concentration.
[Show abstract][Hide abstract] ABSTRACT: Introduction: The development of metastatic liver disease is an adverse prognostic factor in patients with cancer. The normal liver has a dual vascular supply mainly derived from the portal vein, with a smaller contribution from the hepatic artery. The hepatic perfusion index (HPI), which is the ratio of the arterial perfusion to the total hepatic perfusion, is elevated in the presence of liver metastases. HPI can be quantified using slope-based methods [1, 2] which are simpler to compute and easier to implement in clinical settings compared with dual-input kinetic modelling methods. In this study, HPI was evaluated in clinical patients with liver metastases using two slope-based methods – the Miles method  and a modified Blomley method . These were compared with parameters derived using a dual-input single compartment model with population-averaged arterial and portal input functions. The ability to accurately quantify hepatic perfusion is desirable, and may be of clinical value for disease evaluation and assessment of treatment response. Method: 20 DCE-MR datasets (iv. Magnevist® 0.1mmol/kg body weight) of neuroendocrine cancer patients with liver metastases were acquired coronally on a Siemens Avanto 1.5T using a phased array body coil and a 3D FFE sequence. Dynamic data were acquired in pairs during breath-holds on expiration with 5s gap between successive breath-holds. 40 volumes were acquired over a 4 minute period. The imaging parameters were TR/TE = 3.28/1.10 ms, FA = 18°, 12×5 mm slices, NSA = 1, iPAT = 2, FOV = 350 mm 2 , 256×256 matrix. The dynamic scan was preceded by a calibration scan with the same parameters except FA = 2° to allow conversion of dynamic signal intensities to gadolinium concentration. Data analysis was performed using in-house software, MRIW . Dynamic images were registered using a simple rigid body algorithm. The HPI is calculated by dividing the arterial perfusion by the sum of the arterial and the portal perfusion. The two slope-based methods are summarised in figure 1. In both methods, the time of peak splenic concentration, t peak is used as a surrogate to distinguish the arterial and the portal phases in the liver (see figure 1). The modified Blomley method differs from the original  in that the portal component was derived without deconvolving the arterial curve from the liver curve. Peak concentrations from the aorta and the portal vein were measured where available, and the population-averaged values (which were 10.7mM and 3.8mM respectively) were used in the HPI calculation. A dual-input single compartment model was used to analyse HPI using measured population-averaged arterial and portal input functions [4, 5]. ROIs were drawn encompassing the whole liver, the lesion and the surrounding liver. The median HPI values derived from the ROIs were compared using a paired t-test (p-value<0.05 taken to be significant).
[Show abstract][Hide abstract] ABSTRACT: Introduction. Respiratory motion needs managing in most liver MRI protocols. It has been shown that, given a deformation field associated with each group of k-space samples (eg each shot in an interleaved sequence), artefacts in a free-breathing liver image may be corrected  using a general matrix model of motion and sampling scheme . We introduce iDROPS, a model-based method for estimating the deformation fields associated with sub-sampled k-space. Methods. The iDROPS process uses two sampling schemes. Considering a base image of N lines of k-space, the training data is acquired by repeatedly imaging the central N/T lines, giving a low-resolution cine series. The imaging data is acquired in an interleaved fashion: in each frame, every D th line of k-space is acquired, such that the whole of k-space is acquired in D frames. An individual imaging frame, when zero-filled and Fourier-transformed, will exhibit N/D ghosting. Both series are acquired during free breathing. The individual frames of the training series are registered to a chosen exhale reference using a non-rigid fluid algorithm . A breathing model [4,5] is built by linear fit of each component of deformation at each spatial position to a breathing parameter: this can be any quantity (perhaps multi-dimensional) which describes the respiratory cycle, and here we use the mean displacement of the whole image (not just the liver) in the head-foot direction. Hence, given an estimate of the breathing parameter, a displacement field for the whole image may be derived. A breathing parameter is then found for each imaging frame by minimizing a cost function (mutual information or sum of squared differences) between the imaging frame and the model-deformed exhale training frame. The images between which cost is calculated are formed from only those lines of k-space which occur in both: N/(TD) lines, every D th line within the central N/T. Thus these images are low-resolution and ghosted. Deformations, however, must be applied to non-ghosted images: so a model-derived deformation is applied to the complex-valued, partially-reconstructed training frame, which is then returned to k-space, where lines not present in the imaging frame are zeroed. The imaging frame is also part-zeroed to leave only overlapping k-space lines, and both are then reconstructed to ghosted, low-resolution images for comparison. This process is repeated to find a parameter value minimizing the cost function, first by a coarse linear search to find an initial guess, followed by Brent (parabolic-interpolated golden-section) minimization . Finally, the estimated parameters are used to generate deformation fields for each imaging frame. These, along with their iteratively-calculated inverses , are then used in an LSQR-based general matrix motion correction step [1,2]. This approach is readily adapted to handle repeated sampling of k-space and thus to give a corrected image averaged over several frames. The estimated deformations may also be used to reject the most-deformed frames in the imaging series.