Computational fluid dynamics (CFD) and magnetic resonance (MR) gas velocimetry were concurrently performed to study airflow in the same model of human proximal airways. Realistic in vivo-based human airway geometry was segmented from thoracic computed tomography. The three-dimensional numerical description of the airways was used for both generation of a physical airway model using rapid prototyping and mesh generation for CFD simulations. Steady laminar inspiratory experiments (Reynolds number Re = 770) were performed and velocity maps down to the fourth airway generation were extracted from a new velocity mapping technique based on MR velocimetry using hyperpolarized (3)He gas. Full two-dimensional maps of the velocity vector were measured within a few seconds. Numerical simulations were carried out with the experimental flow conditions, and the two sets of data were compared between the two modalities. Flow distributions agreed within 3%. Main and secondary flow velocity intensities were similar, as were velocity convective patterns. This work demonstrates that experimental and numerical gas velocity data can be obtained and compared in the same complex airway geometry. Experiments validated the simulation platform that integrates patient-specific airway reconstruction process from in vivo thoracic scans and velocity field calculation with CFD, hence allowing the results of this numerical tool to be used with confidence in potential clinical applications for lung characterization. Finally, this combined numerical and experimental approach of flow assessment in realistic in vivo-based human airway geometries confirmed the strong dependence of airway flow patterns on local and global geometrical factors, which could contribute to gas mixing.

No abstract

This work reports the use of single-shot spin echo sequences to achieve in vivo diffusion gas measurements and ultrafast imaging of human lungs, in vivo, with hyperpolarized 3 He at 0.1 T. The observed transverse relaxation time of 3 He lasted up to 10 s, which made it possible to use long Carr-Purcell-Meiboom-Gill echo trains. Preliminary NMR studies showed that the resolution of lung images acquired with hyperpolarized 3 He and single-shot sequences is limited to about 6 mm because of the diffusion of the gas in applied field gradients. Ultrafast images of human lungs in normal subjects, achieved in less than 0.4 s with the equivalent of only 130 mol of fully polarized 3 He, are presented. Comparison with other studies shows that there is no SNR penalty by using low fields in the hyperpolarized case. Advantage was taken of the self diffusion-weighting of the rapid acquisition with relaxation enhancement (RARE) sequence to acquire apparent diffusion coefficient (ADC) images of the lungs. Time scales of seconds could be explored for the first time because there is no hindrance from T * 2 as with the usual approaches. At 0.1 T, 180°RF pulses can be repeated every 10 ms without exceeding specific absorption rate limits, which would not be the case for higher fields. Moreover, at low field, susceptibility-induced phenomena are expected to be milder. This supports the idea that low-field imagers can be used for hyperpolarized noble gas MRI of lungs and may be preferred for ADC measurements. Magn Reson Med 47:75-81, 2002.

Magnetic resonance experiments require the main magnetic field, B 0 , to remain very stable. Several external sources, such as moving ferromagnetic objects and/or changing electromagnetic fields, can significantly change the value of B 0 over time. This work describes an apparent displacement along the phase-encoding axis caused by a variation in B 0 . This artifact was observed in fMRI images acquired with EPI. The effect was characterized and tested using an immobile phantom. The image displacement motion along the phase-encoding axis closely followed the changes in B 0 . The stability of the principal magnetic field, B 0 , is critical for MRI experiments. This is especially true for EPI based images, which are very sensitive to phase shifts occurring during long echo train acquisitions, as there is no selfrephasing of the echoes. If the instability occurs with a time scale shorter than the time required to acquire an image, the resulting phase-shift will produce ghost artifacts in the phase-encoding direction (1-3). If it occurs with a longer time scale, the phase-shift will produce image-to-image changes that may be misinterpreted in fMRI experiments as subject head motion or false cortical activation (4). Although most MRI systems are designed to minimize the effect of external sources of magnetic perturbations, moving ferromagnetic objects, such as cars, or changes in electromagnetic fields caused by power lines may have residual effects. The intensity of the effect depends on many parameters, such as the distance between the magnetic or electrical source and the magnet, the design of the magnet and its shielding, and the type of MRI sequence (3,4). This report describes the effect of fluctuations in B 0 produced on a 1.5 T whole-body MRI scanner by a nearby train power line. The artifacts observed in fMRI experiments were characterized using a phantom and corrected with a self-navigator echo (5) algorithm using the k-space center line. THEORYWe refer to the readout, phase-encoding, and slice-selection axes as x, y, and z. Let us consider ⌬B 0 (t), the timedependent variation in the principal magnetic field around its nominal value B 0 (we do not consider the fields produced by the gradient coils). We shall assume that ⌬B 0 (t) is slow enough to be considered as constant during the acquisition of one image. The effect of ⌬B 0 will therefore be a constant and additional phase shift of the signal occurring during the interval between the radiofrequency (RF) pulse to the signal readout:where ␥ is the gyro-magnetic ratio. If y is the time between the beginning of the acquisition of two consecutive lines in the k-space, ⌬B 0 will result in a phase shift ␦ between these lines with:This phase shift is the same as that produced by a displacement ␦y along the y-axis provided that:where k y ϭ ␥ ͵ G y ͑t͒dt. Though a phase shift also occurs along the readout axis, it is negligible in practice. Before such an artifact can be corrected we must know the value of ⌬B 0 for each acquisition, and then compensate for the pha...