Hyperpolarized and Inert Gas MRI: From Technology to Application in Research and Medicine

Hyperpolarized and Inert Gas MRI: Theory and Applications in Research and Medicine is the first comprehensive volume published on HP gas MRI. Since the 1990’s, when HP gas MRI was invented by Dr. Albert and his colleagues, the HP gas MRI field has grown dramatically. The technique has proven to be a useful tool for diagnosis, disease staging, and therapy evaluation for obstructive lung diseases, including asthma, chronic obstructive pulmonary disease (COPD), and cystic fibrosis.

HP gas MRI has also been developed for functional imaging of the brain and is presently being developed for molecular imaging, including molecules associated with lung cancer, breast cancer, and Alzheimer’s disease. Taking into account the ongoing growth of this field and the potential for future clinical applications, the book pulls together the most relevant and cutting-edge research available in HP gas MRI into one resource.

• Presents the most comprehensive, relevant, and accurate information on HP gas MRI
• Co-edited by the co-inventor of HP gas MRI, Dr. Albert, with chapter authors who are the leading experts in their respective sub-disciplines
• Serves as a foundation of understanding of HP gas MRI for researchers and clinicians involved in research, technology development, and clinical use with HP gas MRI
• Covers all hyperpolarized gases, including helium, the gas with which the majority of HP gas MRI has been conducted

• Language: English
• Publisher: Academic Press
• Release date: Nov 17, 2016
• ISBN: 9780128037041

Book preview

1985;27:1–61.

Chapter 1

MRI Acquisition Techniques

J.P. Mugler III

Abstract

Several characteristics of hyperpolarized gases relevant to the design of magnetic resonance acquisitions differ substantially from those for water or lipid protons in biological tissues. Thus, these differences play a critical role in the selection and optimization of pulse sequences for hyperpolarized-gas applications. In this chapter, the general impact of these characteristics on the design of pulse sequences for magnetic resonance imaging of hyperpolarized ³He or ¹²⁹Xe is described. Since the majority of research in this field has involved applications in the lung, the discussion focuses on concepts applicable to pulmonary imaging, particularly in humans. Main topics include the nonequilibrium nature of hyperpolarized magnetization, high diffusivity of gases, design considerations for gradient-echo pulse sequences, and the role of ¹²⁹Xe gas exchange and uptake in the lung.

Keywords

Hyperpolarized gas; pulse sequences; ³He; ¹²⁹Xe; pulmonary imaging; diffusion; variable flip angles

Several characteristics of the hyperpolarized gases ³He and ¹²⁹Xe, relevant to magnetic resonance acquisitions, differ substantially from those for water or lipid protons in biological tissues. Thus, these differences play a critical role in the selection and optimization of pulse sequences for hyperpolarized-gas applications. Compared to water or lipid protons in biological tissue, important characteristics of these gases include:

1. Hyperpolarized longitudinal magnetization is not at thermal equilibrium in the static magnetic field of the MR scanner;

2. Diffusivity is many times higher;

3. Relaxation times are typically much longer, and the longitudinal relaxation rate is directly proportional to the concentration of oxygen; and

4. For ¹²⁹Xe in certain environments, such as the lung, states with substantially different chemical shifts are in dynamic exchange.

In this chapter, the general impact of these characteristics on the design of pulse sequences for MR imaging of hyperpolarized ³He or ¹²⁹Xe is discussed; details associated with pulse sequences for specific applications are described in subsequent chapters. Since the majority of research in this field has involved applications in the lung, the discussion focuses on concepts applicable to pulmonary imaging, particularly in humans.

Nonequilibrium Magnetization

By definition, hyperpolarized longitudinal magnetization is not at thermal equilibrium, which has a profound effect on the design and use of pulse sequences for hyperpolarized-gas imaging [1–8]. For example, standard scanner-calibration pulse sequences, designed for thermal-equilibrium magnetization, as well as widely used pulse sequences, such as spin-echo imaging or turbo/fast spin-echo imaging, are generally not appropriate.

Pulse sequences used for proton MRI generally rely on recovery of the longitudinal component of the magnetization (MZ) via T1 relaxation to achieve adequate signal levels in the images. However, because hyperpolarized magnetization is not at thermal equilibrium in the static magnetic field of the MR scanner, and because it is several orders of magnitude larger than the corresponding thermal-equilibrium magnetization, T1 relaxation results in decay, rather than regrowth, of longitudinal magnetization.

This characteristic difference between thermally polarized and hyperpolarized magnetization is illustrated for a single spin-echo pulse sequence in Fig. 1.1A, which shows the temporal evolution of MZ for thermally polarized and hyperpolarized magnetization during the first three repetitions of the pulse sequence. Thermally polarized magnetization exhibits the familiar behavior that, following each 90°–180° pair of radio-frequency (RF) pulses, MZ recovers, due to T1 relaxation, to a substantial fraction of its initial value (thermal equilibrium, Mo) during the time period between applications of the RF-pulse pairs. In contrast, the hyperpolarized MZ does not relax back toward its initial value (MH) following the RF-pulse pairs. Instead, following the first 90° RF pulse, it relaxes back toward thermal equilibrium (Mo), which is many times less than MH, and, as a result, the hyperpolarized magnetization decays due to T1 relaxation rather than recovering. The practical consequence is that the hyperpolarized MZ remains essentially zero following the first 90° RF pulse. (Since the density of gases is roughly a thousand times less than that of soft tissue, the thermal-equilibrium signal from helium or xenon gas is far too small to be useful for practical in vivo applications, although pressurized, thermally polarized gas samples can be useful for quality control measurements or pulse-sequence development.) Thus, pulse sequences that require multiple, high flip angle excitation RF pulses, such as spin-echo or turbo/fast spin-echo, are generally not useful for imaging hyperpolarized gases. An exception is when additional hyperpolarized gas can be introduced between excitations, which may be practical for imaging a small animal, such as a mouse, but is typically not practical for imaging humans because of the volume of hyperpolarized gas that would be required.

Figure 1.1 Temporal evolution of the longitudinal component (MZ) of thermally polarized and hyperpolarized magnetization during (A) the first three repetitions of a single spin-echo pulse sequence and (B) 120 repetitions of a gradient-echo pulse sequence. The initial values for the thermally polarized and hyperpolarized magnetizations are Mo and MH, respectively. The MZ curves were calculated using the following parameter values: repetition time, 800 ms (spin echo) or 20 ms (gradient echo); echo time, 50 ms (spin echo); T1, 1 s (thermal) or 20 s (hyperpolarized). Ideal spoiling of transverse magnetization between pulse-sequence repetitions was assumed.

Considering that the total magnetization associated with an inhalation of hyperpolarized gas is fixed, an effective acquisition approach entails consuming only a small portion of the hyperpolarized magnetization with each excitation, as is done by a low flip angle gradient-echo (GRE) pulse sequence. Fig. 1.1B illustrates the temporal evolution of MZ for thermally polarized and hyperpolarized magnetization during such a pulse sequence. As is well known, thermally polarized MZ begins at Mo and gradually approaches a steady-state value that is reached when the fraction of MZ consumed by each excitation RF pulse is balanced by T1-dependent regrowth of MZ between RF pulses. Since analogous recovery of MZ does not occur for a hyperpolarized gas as explained above, it is fortunate that the T1 relaxation time for ³He or ¹²⁹Xe in the lung is rather long (~20 s), so that the decrease in MZ from T1 decay is not substantial over a period of a few seconds. Although the hyperpolarized MZ decreases monotonically as the low flip angle excitation RF pulses are applied, an appropriately chosen flip angle value (as discussed later in the chapter) provides MZ values during the acquisition that are similar to those for thermally polarized magnetization, as shown in Fig. 1.1B. Due to this favorable behavior for a low flip angle GRE acquisition applied to nonequilibrium magnetization, they are used in the majority of hyperpolarized-gas applications.

High Diffusivity

The self-diffusion coefficients of ³He (~1.8 cm²/s [9]) and ¹²⁹Xe (~0.06 cm²/s [10]) gases are several orders of magnitude larger than that for water in biological tissues and, as a result, even the magnetic field gradients used for spatial encoding can cause substantial signal attenuation. There are both positive and negative aspects of this situation. On the positive side, one can take advantage of diffusion-induced signal attenuation to quantitatively characterize porous structures, such as the lung, using hyperpolarized gases. On the negative side, diffusion-induced signal attenuation can limit spatial resolution for echo-train pulse sequences and lower the signal-to-noise ratio (SNR) for a variety of pulse-sequence types. First, the negative consequences of diffusion-induced signal attenuation are explored, and then pulse-sequence configurations commonly used for diffusion-weighted imaging are described.

Signal Loss From Diffusion

Assuming isotropic, unrestricted diffusion, it is straightforward to calculate the diffusion-induced signal attenuations corresponding to various pulse-sequence types. However, the lung is a complex, restricted-diffusion environment for which it has been shown that the apparent diffusion coefficient (ADC) depends on the timing, direction, and strength of diffusion-sensitization gradients [11–19]. Nonetheless, by using the mathematical formalism appropriate for isotropic unrestricted diffusion, by restricting attention to a specific time scale (typically a few milliseconds), and by choosing ADC values that are representative of this chosen timescale and the status of the lung tissue, signal-attenuation values that approximate those obtained in the lung can be calculated, thus providing useful guidance for the design of pulse sequences for imaging the lung. Even so, one should keep in mind the limitations of this approach when applying the associated results to any particular application of interest.

As noted above in section Nonequilibrium Magnetization, GRE pulse sequences are commonly used for hyperpolarized-gas imaging. For parameter selections associated with many applications of practical interest, the readout (frequency-encoding) gradient is the dominant source of diffusion-induced signal attenuation for this type of pulse sequence [20]. Assuming an idealized bipolar waveform for the readout gradient (i.e., ramp times equal to zero, equal gradient magnitudes for the dephasing and readout portions of the waveform, and symmetric sampling of the echo; see Fig. 1.2A), a very simple expression for the b value at the echo time can be derived:

(1.1)

where Δx is the nominal spatial resolution along the readout direction and Ts is the duration of data sampling for each echo. (The b value is a quantity associated with a gradient waveform that reflects the signal attenuation resulting from diffusion during application of the waveform.) Signal attenuation at the echo time, calculated based on Eq. (1.1) as e−b(ADC), is shown in Fig. 1.2B for relatively short (1 ms) and long (10 ms) data sampling periods (corresponding to receiver bandwidths of 1000 and 100 Hz/pixel, respectively), assuming ³He and ¹²⁹Xe ADC values associated with these gases in the healthy human lung. For spatial resolutions used in human imaging (typically 2–5 mm) the signal loss is no more than 20%, however for higher resolutions, as would be appropriate for small-animal studies, the signal loss becomes significant.

Figure 1.2 (A) Timing diagram for an idealized readout gradient in a GRE pulse sequence (GR, readout gradient; TS, data sampling period). (B) Signal attenuation at the echo time of a GRE pulse sequence, as induced due to diffusion along the direction of the readout gradient shown in (A), versus nominal spatial resolution for ³He and ¹²⁹Xe gases in the healthy human lung (ADC: 0.2 cm²/s for ³He and 0.04 cm²/s for ¹²⁹Xe) and a data sampling period of 1 or 10 ms.

Diffusion-induced signal attenuation at the echo time can be reduced or eliminated by using asymmetric sampling of the echo [20], spiral sampling [21], or radial sampling [22], but diffusion-induced blurring during the readout period nonetheless remains. Note that, considering the gradient-amplitude and slew-rate limits of a particular scanner, the actual diffusion-induced signal attenuation at the echo time will differ from values estimated using Eq. (1.1).

From the perspective of nonequilibrium magnetization, single-shot echo-planar imaging would appear to be a candidate pulse sequence for hyperpolarized-gas imaging since only one excitation RF pulse is required per image. An echo-planar readout is composed of concatenated bipolar waveforms (i.e., just a series of GRE readouts). Thus, the associated signal attenuation from the readout gradient at the echo time is estimated as SN, where S is the signal attenuation for a single bipolar waveform and N is the number of bipolar waveforms applied before the center of k space, along the phase-encoding direction, is sampled. For example, S for ³He, 1-ms data sampling period and 2-mm resolution is 0.984, so the signal attenuation for an echo time occurring after 32 bipolar waveforms is 0.984³²=0.60 (40% signal loss). Thus, single-shot echo-planar imaging is associated with substantial signal loss due to diffusion, unless the spatial resolution is relatively low [23].

Similarly, single-shot spin-echo-train imaging (e.g., single-shot turbo/fast spin echo) would also appear to be a candidate pulse sequence. To understand how diffusion-induced signal attenuation for the readout gradient of this pulse sequence compares to that for a GRE pulse sequence, it is useful to recall that the b value associated with a gradient waveform can be calculated as the integral of the squared magnitude of the k-space trajectory [24]. This relationship is extremely useful as an intuitive guide for the design of pulse sequences. For example, Fig. 1.3 compares the GRE readout (Fig. 1.3A, plotted up to the echo time) to the standard (monopolar) readout gradient waveform typically used in a RARE-type [25] (RARE, Rapid Acquisition with Relaxation Enhancement) spin-echo-train pulse sequence (Fig. 1.3B). The bipolar nature of the GRE readout results in a much lower b value than that for the RARE readout. That is, the area under the |k|² curve, shown in gray in Fig. 1.3A, is much smaller than that in Fig. 1.3B because, for the RARE case, the k-space trajectory remains at a high spatial frequency between the end of a given readout and the beginning of the subsequent readout. Diffusion-induced signal attenuation during the echo train ultimately limits the spatial resolution by widening the point spread function and also decreases the SNR for the central portion of k space unless centric phase encoding is used. As discussed in the context of GRE pulse sequences, diffusion-induced signal attenuation can be maintained at a tolerable level by using relatively low spatial resolution. For example, in a ³He lung imaging study using a RARE-type pulse sequence [26], a minimum resolution of 6 mm was postulated based on an attenuation limit of 37% for the signal remaining at the end of a 36-echo train.

Figure 1.3 Analysis of the diffusion sensitivity associated with the readout gradient for simplified versions of the following pulse sequences: (A) GRE (B) RARE; and (C) RARE with balanced readout gradient (180° RF pulse) or balanced steady-state free precession (α° RF pulse). The RF timing, readout gradient, and associated squared magnitude of the k-space trajectory (|k|²) are shown from the excitation RF pulse to the echo time for (A), and from a given echo time to the subsequent echo time for (B) and (C). The diagrams assume equal spatial resolution and data sampling periods for the three cases. The b value associated with the readout gradient is the area under the |k|² curve.

Improvements in the performance of spin-echo-train pulse sequences can be obtained by optimizing the gradient waveforms. For example, Fig. 1.3C illustrates that by modifying the readout waveform for the spin-echo-train pulse sequence from the standard monopolar form to a balanced readout, the b value (area under the |k|² curve) is reduced to less than half of that for the monopolar waveform (Fig. 1.3B), even though the gradient is active for twice as long for the balanced waveform, providing a corresponding reduction in diffusion-induced signal attenuation. This occurs because the trajectory corresponding to the first portion of the balanced readout returns to the center of k space before the subsequent RF pulse. Fig. 1.4 shows the theoretically predicted signal versus echo number for a balanced readout gradient. Despite the improvements associated with the balanced readout, there remains substantial signal loss by the end of a moderate-length echo train, even without considering the concurrent effects of T2 relaxation. Nonetheless, this analysis of the behavior associated with different readout waveforms has practical utility because the same balanced gradient waveform is used in another potentially valuable pulse sequence for hyperpolarized-gas imaging—balanced steady-state free precession (bSSFP [27–34]). (For simplicity, the term bSSFP will be used to reflect the pulse-sequence structure under consideration, even though a steady state is never reached when imaging hyperpolarized magnetization.) Compared to a RARE-type pulse sequence, the bSSFP pulse sequence can provide superior behavior in the face of diffusion-induced signal loss by using flip angles much lower than 180° so that only a small fraction of the available magnetization is in the transverse plane and therefore subject to diffusion-induced signal loss between succeeding echoes, whereas in the RARE pulse sequence (using 180° refocusing RF pulses) all of the magnetization is in the transverse plane. Diffusion-induced signal behavior and other aspects of bSSFP pulse sequences associated with hyperpolarized-gas imaging of the lung are discussed later in the chapter. For a more general discussion of diffusion effects in echo-train pulse sequences, including, e.g., the case shown in Fig. 1.3B but with a flip angle other than 180°, the reader is referred to reference [35].

Figure 1.4 Theoretical calculations of normalized signal versus echo number for a spin-echo-train pulse sequence using a balanced readout gradient and 180° refocusing RF pulses, illustrating the diffusion-induced signal attenuations for nominal spatial resolutions of 2, 4, and 6 mm that result from an ADC of 0.2 cm² (³He in a healthy human lung). The calculation included the signal-attenuation effects of a centrically ordered phase-encoding gradient.

Diffusion-Weighted Imaging

Diffusion-weighted imaging using hyperpolarized gases has shown great potential for quantitatively characterizing lung structure. For example, ADC measurements are sensitive to disease-related changes in lung microstructure [11,36–42] and may permit detection of subclinical changes in microstructure before they are seen on high-resolution computed tomography [43]. In addition, ADC measurements reveal subtle gravity-dependent gradients in alveolar size [41,44–47], as well as age-related changes [48,49]. Details concerning the application and interpretation of diffusion measurements are discussed in subsequent chapters, particularly Chapter 12, Lung Morphometry With HP Gas Diffusion MRI. Here, the most commonly used pulse sequences for diffusion-weighted imaging of hyperpolarized gases are briefly described.

The human lung is a complex, restricted-diffusion environment with characteristic length scales ranging from ~200 μm (approximate diameter of an alveolus) to ~2 cm (approximate diameter of the trachea). The diffusion time determines the extent of lung structure explored by diffusing gas atoms during the measurement. For example, for freely diffusing ³He in air, the root-mean-squared (RMS) diffusion distances range from ~230 μm (approximate size of an alveolus) to ~7 cm (greater than the size of the trachea) for diffusion times from 0.1 ms to 10 s (see Fig. 1.5), and these distances are correspondingly reduced for the ADC values associated with restricted diffusion in lung tissue. Thus, diffusion times much greater than a few milliseconds are required for hyperpolarized-gas atoms to explore connectivity among the smallest generations of airways (e.g., acini; see Fig. 1.5) as well as potential collateral-ventilation pathways [50–53]. To permit different length scales in the lung to be probed, pulse sequences have been developed with diffusion times on the order of a millisecond—so-called short-timescale measurements—and with diffusion times ranging from many milliseconds to several seconds—so-called long-timescale measurements. Although long-timescale ADC measurements may offer improved sensitivity for detection of pathological changes in airspace connectivity [15,17,54,55], and may thus provide fundamentally different information than that from short-timescale measurements, the vast majority of hyperpolarized-gas diffusion studies have focused on short-timescale measurements due to their relative simplicity and robustness.

Figure 1.5 Relationship between diffusion time and the root-mean-squared (RMS) distance that dilute ¹²⁹Xe or ³He gas atoms in air diffuse in an unrestricted environment (diffusivity approximately 0.14 or 0.86 cm²/s, respectively [57]). The dashed (light blue) lines indicate the diffusion distance difference between ¹²⁹Xe and ³He due to the difference in diffusivity of the two gases. Approximate dimensions for several structures in the human lung are shown for comparison to the ³He diffusion distance. Source: Adapted from Fig. 3 in Mugler JP, III, Wang C, Miller GW, Cates GD, Jr, Mata JF, Brookeman JR, et al. Helium-3 diffusion MR imaging of the human lung over multiple time scales. Acad Radiol