Open access peer-reviewed chapter

# High Resolution 3He Pulmonary MRI

By Matthew S. Fox and Alexei V. Ouriadov

Submitted: October 1st 2018Reviewed: January 28th 2019Published: March 6th 2019

DOI: 10.5772/intechopen.84756

## Abstract

Hyperpolarized gas MRI of the mouse lung is of great interest due to the urgent need for novel biomarkers for the assessment of respiratory-disease progression and development of new therapies. Small animal 3He lung MRI requires high-spatial-resolution imaging (<500 μm) to obtain acceptable images for visualization of all branches of lung microstructure from the mouse trachea to lung parenchyma. The use of conventional fast-gradient-recalled echo (FGRE) pulse sequences for high-spatial-resolution mouse lung imaging is challenging due to the signal loss caused by significant diffusion-weighting by the imaging gradients, particularly in larger airways where 3He diffusion is maximized. In this chapter, a modified FGRE approach called X-Centric is described for acquiring 3He mouse lung MRI. X-Centric is a center-out technique, allowing high-spatial-resolution, and high signal-to-noise ratio density-weighted imaging, as it is a short-TE method minimizing diffusion decay. Here, we also take advantage of a high-performance insertable-gradient-set interfaced with a clinical MRI system and a custom-built constant-volume ventilator to get the maximum benefits of X-Centric. High-spatial-resolution 3He X-Centric imaging was performed in a phantom and mouse lungs, yielding a nominal resolution of 39 μm and 78 μm respectively. We also demonstrate the feasibility of 129Xe/19F X-Centric MRI in a phantom and in rat lungs.

### Keywords

• X-Centric
• 3He
• 129X
• 19F mouse
• rat
• lung
• short-TE
• x-nuclei
• pre-clinical
• hyperpolarized
• MRI
• COPD
• CF
• BPD
• alpha-1 antitrypsin deficiency

## 1. Introduction

### Table 1.

Comparison of 3He, 129Xe, inert fluorinated gas and propane.

SF6 and 3He T2* values shown within major airwaya and parenchemab of the rodent lung. 129Xe T2* values shown for xenon gasa within the major airway and xenon dissolvedc within lung tissue (barrier) and red blood cells. 129Xe and 3He T1 valuesd measured in rodent lungs after several wash-out (anoxic) breaths of the hyperpolarized gas. Cost of the 86% enriched 129Xee.

### 2.2 Relaxation times

Generally, both longitudinal and apparent transverse relaxation [44] times (T1 and T2*) of 3He are not physical constants and they both depend on the main magnetic field strength, its gradients, presence of paramagnetic impurities interacting with the hyperpolarized gas (e.g., oxygen), and finally, diffusivity. Normally, the main mechanism for T1 decay of hyperpolarized gases (i.e., polarization losses in contrast to conventional thermally polarized spins) is the presence of oxygen in lung [45]. The T1 decay rate is inversely proportional to the partial oxygen pressure (pO2) [46]:

1T1=1T1,0+κ0timepO2tdtE1

where T1,0 is the longitudinal relaxation time expected in the absence of O2 and κis the relaxivity of 3He due to the presence of O2 (0.38 atm−1 s−1 at body temperature [46]).

This suggests that a small number of anoxic pre-breaths or wash-out breaths prior the actual 3He MRI measurements can be helpful for minimizing the oxygen concentration in lung prior to acquisition and administration of hyperpolarized gas. Table 1 suggests that the T1 in small animal lungs after applying a number of wash-out breaths can be sufficiently long [47] compared to the typical image acquisition times of 1 s [27] for mice and 10 s for rats [37]. As previous work shows, eight wash-out breaths drastically reduce the O2 concentration [17, 23, 48]. The value of T2* for 3He normally varies between the major airways (1.5 ± 0.25 ms [37]) and lung parenchyma (3.0 ± 1.0 ms [37]) due to the differences in airway size which have a significant effect on the mode of diffusion (e.g., free diffusion vs. restricted) and amount of signal loss due to diffusion. Table 1 shows T2* values obtained for hyperpolarized gas within rat lungs at 3 Tesla. As expected T2* estimates obtained for the major airways were smaller than the estimates observed in lung parenchyma.

### 2.3 Signal to noise ratio

An imaging sequence should be optimized to simultaneously ensure maximum SNR and minimal blurring in order to successfully perform high-resolution hyperpolarized 3He MR measurements. Understanding of the mechanisms that affect SNR and contribute to image blurring is the basis for the developing an imaging technique appropriate for hyperpolarized gas imaging. Generally, an expression for the SNR of a FGRE sequence can be expressed as [42]:

where ΔxΔyΔz is the voxel size, NxNyNz is the number of k-space samples in each direction, BWread readout bandwidth and fρPαT2T1bDis the pulse-sequence-dependent function that determines the signal amplitudes at readout for the center of k-space [49]:

fρPT1T2αbDρPsinαexpTE/T2expTR/T1expbDE3

As it can be seen, the signal amplitudes at readout depend on the density (ρ) and polarization (P) of 3He, flip angle (α), TE, repetition time (TR) and finally, diffusion attenuation [50] introduced by slice select and/or frequency encoding gradients. In order to simultaneously minimize T1 decay and T2*-induced blurring and maximize signal, one should use a short-TE (TE < 1 ms) center-out k-space traversal pulse sequence to ensure only density-weighting at the center of k-space. For the case of TE < T2*, and if the total acquisition time (1–10 s) is much less than T1 (25–90 s) the pulse sequence dependent function can be expressed as:

fρPαbDρPsinαexpbDE4

where bis the b-value, the diffusion weighting factor that depends on the diffusion gradient duration and magnitude (which in this case are imaging gradients); and Dis the diffusion coefficient of 3He in the lung (0.1–2 cm2/s) [37, 50].

### 2.4 Flip angle

Let us consider each factor in Eq. (4) individually in order to optimize this function. As the hyperpolarized magnetization is not renewable due to the its non-equilibrium nature, it needs to be spent (flipped by RF pulses) very efficiently. A single 90o RF pulse can effectively waste all of the available magnetization. We start with the two approaches of setting flip angles in hyperpolarized gas MRI. The first method is the Constant Flip Angle (CFA) approach [51] or small flip angle approach (1o < flip angle <10o). Using a CFA approach, the optimal flip angle (αopt), which provides the highest signal for sequential phase encode ordering, is expressed as [51]:

αopt=tan12/NyNzE5

Unfortunately, the use of the optimal flip angle given by Eq. (5) leads to significant signal decay due to the constant-value RF pulses applied during k-space acquisition. As an example, for a 2D case where Ny = 128 and αopt = 7.1o the signal decays by ∼60% across the k-space). An undesirable consequence of such decay is reduced image resolution as result of the RF pulse “history” during k-space acquisition. Point-spread-function (PSF) simulations for a 2D case with center-out phase encode ordering (Figure 1) confirmed that the CFA approach leads to image blurring (Full Width at Half Maximum (FWHM) = 1.5 pixels) [47]. This signal decay-induced image blurring can be eliminated by using a Variable Flip Angle (VFA) approach (1o < flip angle <90o) [52]. VFA mitigates signal loss during image acquisition (i.e., blurring) by ensuring a constant signal for each phase-encode line by starting with a low flip angle and incrementing the flip angle of each RF pulse, line after line [52]. The flip angle for each ith phase-encode line of a VFA sequence can be calculated from the following equation [47]:

αi=tan1expNyNziTR/T1NyNziE6

For the case where the total image acquisition time is much less than T1 (T1 typically >10 sec under normal conditions), Eq. (6) takes the simpler form [47]:

αitan11NyNziE7

For a 2D case (Ny = 128), α1 = 5.1o and α128 = 90o, producing no CFA-like signal decay, and consequently, no image blurring due to RF pulse “history” [47]. Simulations have shown that VFA only shows blurring of 1.2 pixels due to discrete sampling [47] and because of these benefits this method is preferable for high spatial resolution 3He imaging.

### 2.5 b-Value

Eq. (4) suggests that signal decay due to diffusion is the main reason the MR signal degradation (assuming minimal T1 and T2* induced blurring and signal loss) and consequently poses as the main difficulty for high spatial resolution 3He imaging. Figure 2 shows the calculated NMR signal as a function of b-value for two different diffusion coefficients, typical for a small animal airways and lung parenchyma respectively [37]. It can be seen that for a b-value >3.5 s/cm2, the signal intensity in the airway is almost zero and there is a 60% loss of signal for lung parenchyma. Let us consider the dependence of b-value on the diffusion gradient duration and magnitude. In general, the b-value for any gradient waveforms can be calculated from the first principles [42]:

b=γ20t10tGtdt2dtE8

where γ is gyromagnetic ratio, Gis gradient amplitude and t, t1 is gradient duration. Unfortunately, Eq. (8) cannot be analytically simplified further for a gradient waveform specific to the frequency encoding imaging gradient. For a relatively simple diffusion gradient waveform such as two bipolar rectangular pulses, the b-value can be calculated based on the following equation [53]:

b=23γ2G2δ2tdwithtd=Δδ3,E9

where δ is the diffusion gradient duration, Δ is a distance between two gradient pulses and td is the diffusion observation time. Note, that b-value strongly depends on the gradient duration (δ2td = time3) then the magnitude (G2), which suggests that minimizing the gradient duration reduces the b-value and consequently minimizes the signal loss. This is especially critical for high-resolution 3He imaging, as the imaging gradient magnitude cannot be really decreased when trying to achieve high spatial resolution, or small FOV for mice.

### 2.6 X-centric

The X-Centric approach requires two discrete acquisitions for each k-space line making the method twice as long in acquisition time compared to a conventional partial-echo FGRE approach for a chosen spatial resolution and matrix size [17, 27, 39]. Though, this is not a significant cost to achieve high-resolution imaging of the lungs of small animals, the scan time can be further reduced by using a partial-echo method in the phase-encoding direction [17, 27, 39]. Note, that the use of a partial-echo method in the phase-encoding direction makes the X-Centric scan time only 10% longer than the partial-echo FGRE sequence for a similar spatial resolution and matrix size [27].

### 2.7 Diffusion attenuation

The exact b-value for gradient waveforms presented in Figure 4 can be calculated from Eq. (8). Thus, b-values calculated for the partial-echo and X-Centric FGRE (256 × 256 matrix size) were 2.29 and 0.061 s/cm2 respectively. Figure 2 indicates that a b-value more than 2.0 s/cm2 leads to the complete signal decay in the major airway which is a significant limitation for high spatial resolution 3He imaging of the mouse lung. In contrast, the X-Centric sequence should theoretically ensure minimal signal decay for given diffusion coefficients. Note that for the unrestricted 3He diffusion case the theoretical SNR (SNRtheor) should depend only on b-value and D:

SNRtheor=S0expbDE10

where S0 is the initial signal (S0 = 1), Dis the free 3He self-diffusion coefficient (2.0 cm2/s) and btakes on a value determined by the chosen FOV and BWread.

According to Eq. (2), for the case of the high-resolution 2D 3He phantom imaging, the SNR should generally depend on the diffusion attention, image resolution, BWread and the first flip angle in the VFA scheme [27]:

as Δx, Δy can be expressed through FOVs and Nx, and Ny Eq. (11) can be rewritten as

If one keeps FOV, BWread, 3He polarization and volume constant across all phantom measurements, then the SNR comparison between the partial-echo and X-Centric FGRE can be done by normalizing the experimental SNR by the first flip angle in the VFA scheme (α1) and matrix size (NxNy). Thus, for a SNR normalized by the respective first applied flip angle and matrix size, one can write the following expression [27]:

SNRnor=SNRsinα1NxNy=Const·expbDE13

where Const=FOVxFOVyBWread. This final SNRnor equation can be used for SNR comparisons of the phantom images obtained for different resolutions with the partial-echo and X-Centric FGRE methods; also this equation is similar to Eq. (10).

### 2.8 Phantom imaging

A plastic 10 mL syringe filled with hyperpolarized 3He was used to validate the X-Centric sequence for high spatial resolution imaging [27]. Figure 5 shows a 2D coronal view whole projection phantom images obtained for partial-echo (top panel) and X-Centric (bottom panel) FGRE for five different resolutions. Images start from 64 × 64 matrix size and end with 512 × 512 from left to the right [27]. The images indicate that at low nominal resolution (625 μm, 64 × 64 matrix size) there is no qualitative or visual difference in the SNR between the images obtained for the partial-echo and X-Centric FGRE. However, there is virtually no signal in phantom images obtained with partial-echo FGRE when nominal resolution is greater than 256 × 256 (78 μm nominal resolution). As expected, the signal intensity of the images obtained with X-Centric gradually decays with increasing nominal resolution and decreasing flip angle (because the first pulse for VFA is a function of resolution, Eq. (7)) but even for 512 × 512 matrix size (39 μm nominal resolution) it does not go below the noise floor as is seen in the case of the partial-echo FGRE. The SNRnor dependence on resolution (bottom axis) and/or b-value (top axis, partial-echo case only) for the phantom images obtained with the partial-echo (solid squares and solid line) and X-Centric FGRE (solid circles and solid line) is plotted in Figure 6. Clearly, the SNRnor obtained for the X-Centric did not depend on image resolution, i.e., diffusion attenuation was minimal even for very high image resolution (39 μm). The experimental phantom results were consistent with the predicted SNR (SNRtheor) calculated based on Eq. (10) for the partial-echo (open squares and dashed line) and X-Centric FGRE (open circles and dashed line). Eqs. (10) and (13) indicate that for the case of unrestricted 3He diffusion (D = 2 cm2/s) one should observe minor signal loss (∼10%) using X-Centric and almost 100% signal loss using the partial-echo FGRE for a nominal resolution greater than 78 μm. Thus our consideration of the factors determining the SNR for high spatial resolution 3He MR measurements shows that in the case of relatively short TR and TE (pulse sequence parameters for 256 × 256 matrix size are present in Table 2) diffusion is the major reason for signal loss. Using the X-Centric method, diffusion decay due to the x-gradients is significantly minimized. It should be noted that we observed a slight decrease of the SNRnor (around 10%) for a nominal resolutions better than 75 μm (Figure 6) which were likely due to T2* decay because as resolution increases, TE becomes longer due to the longer duration of phase-encoding gradient. Nevertheless, X-Centric phantom images show no evidence of image blurring due to the RF pulse history or T2*-based signal decay in the frequency/phase encoding directions. The phantom measurements have confirmed that significant diffusion attenuation induced by the x-gradient with the partial-echo FGRE sequence can be the main restriction for high spatial resolution 3He imaging of the lung.

3He MRI129Xe MRI19F MRI
TE/TR (ms)0.70/3.70.81/50.0.50/4.0
Flip angle3.2o (VFA)6.4o (CFA)70 o (EA)
FOV (mm2)20 × 20250 × 25060 × 60
Matrix size256 × 160128 × 8064 × 40
Number of averages1160
Slice orientationAxial/coronalAxial/coronalAxial/coronal
Pixel size (mm3)0.078 × 0.0781.9 × 1.9 × 3000.94 × 0.84
Scan time (s)1.1>8>10
MRI scannerGE 3 T MR750Philip. 3 T AchievaPhilip. 3 T Achieva

### Table 2.

MRI acquisition parameters for methods in this chapter.

VFA = variable flip angle; CFA = constant flipped angle; EA = Ernst angle; FOV = field of view; TR = relaxation time; TE = echo-time.

## 3. High spatial resolution imaging of mouse lung

### 3.1 Animal preparation

The animal protocol was approved by the Animal Use Subcommittee of the University Council on Animal Care at the University of Western Ontario, London, Ontario, Canada and the Animal Care Committee at Merck Frosst Canada, Kirkland, Quebec, Canada. Each mouse (BALB/c) was pretreated with Midazolam (1 mg/kg) intraperitoneally (i.p.) 5 min before anesthesia with ketamine (95 mg/kg, i.p.) and xylazine (6.4 mg/kg, i.p.). Following anesthesia, the tail vein was cannulated (26 GA Abbott catheter) to maintain anesthesia with ketamine (30 mg/kg) and xylazine (2 mg/kg) every 40 min. To allow artificial mechanical ventilation, the trachea was cannulated by tracheotomy using an 18 GA Teflon catheter. A line for administration of pancuronium (0.8 mg/kg, i.p.) was established using a 26 GA cannula to allow for injections after the mouse was being ventilated inside the scanner. Other physiological instrumentation included ECG and a rectal temperature probe (SA Instruments Inc., Stony Brook, NY).

### 3.2 Ventilation

It was a very challenging task to design a ventilator to ventilate mice inside an MRI because of the restrictions due to the high magnetic field environment and the requirement to accurately deliver very small tidal volumes of gas to the lungs of mice at a high ventilation rate. Under spontaneous breathing, the mouse maintains its oxygenation with a tidal volume less than 0.1 ml [54, 55]. Because of the dead space volume within the tubing necessary for connecting mouse airways to the ventilator, the mouse was normally ventilated with tidal volume of 0.2 ml (8 ml/kg). In this range, the amount of the ventilating gas (or fresh gas) getting compressed in the gas line could significantly reduce the portion of fresh gas entering the lungs for gas exchange. Without a proper correction for this effect, adequate ventilation would not be achieved at this ventilating setting. In many cases, the ventilation was set with a much larger tidal volume (>10 ml/kg) at a slower ventilation rate (60–90 bpm). Since lung volume under spontaneous breathing is only 0.3–0.4 ml [56], the larger tidal volume approach could overextend the airways. As a result, many mice were observed to expire shortly after the ventilation, impeding the completion of studies.

To resolve these problems, we modified a flexiVent ventilator (Figure 7, Scireq, Montreal, QC, Canada) that incorporated real-time monitoring of volume and pressure to allow corrections of the gas compression effect. To minimize the gas line volume and ensure the effectiveness of ventilation, a non-metal swing valve was designed and placed close to the animal (inside the scanner) to regulate the gas flow. Different than the pneumatic valves normally used in the magnetic environment, the swing valve was connected to two driving solenoid valves (which reside far away from the high magnetic field) via two multiple-section light-weight carbon fiber rods. This mechanical design enabled fast open/close activations. The activation of a solenoid valve pulls the swing valve to close the ports on one side and open the ports on the other side. Two solenoid valves were activated in sequence to generate a swing motion that opens and closes the ports on the respective sides alternately. This swing motion was synchronized with the plunging motion of the piston that drives the gas into the lung to form a ventilation cycle. Because of the fast response of the swing valve, the modified flexiVent ventilator is capable of performing the forced oscillation method to access pulmonary mechanics. To minimize oxygen-induced depolarization (T1 decay) of hyperpolarized 3He gas, the flexiVent ventilator was modified to deliver breathing gas and hyperpolarized 3He gas separately using two individual syringes. The ventilation system was integrated to the MR scanner (Figure 7) using custom-designed LabView software (National Instruments) to allow ventilation-synchronous acquisition. Following the animal preparation, the animal was connected to the ventilator and ventilated using a tidal volume of 8 ml/kg at 100 bpm and was placed inside the scanner. The breath-hold time for imaging was 1 s.

### 3.3 3He polarization and delivery

3He gas was polarized using an optical pumping spin-exchange system (Helispin™, GE Healthcare, Durham, NC). The 3He gas was typically polarized for at least 8 hours to achieve a polarization over 35%. Prior to transferring the hyperpolarized 3He gas to a 150 mL Tedlar bag (Jensen Inert Products, Coral Springs, FL), the bag was carefully washed with N2 gas three times to eliminate any oxygen contamination. Following the gas filling, the 3He bag was placed inside a custom-made plexiglass chamber and connected to a 3He intake line of the ventilator. The chamber was then placed within the homogenous B0 field inside the scanner to conserve 3He polarization (T1 = ∼43 min). To facilitate gas intake by the ventilator, the pressure in the chamber was kept at 3 cm-H2O during the gas delivery.

To validate the X-Centric imaging method, 2 mL at 1 atm of the hyperpolarized 3He gas was drawn into a 10 mL syringe for phantom testing [27].

### 3.4 MRI hardware and sequence parameters

Hyperpolarized 3He MR imaging was performed on a GE clinical scanner (3 T, Excite 12.0) which was converted using a home-made insert gradient set (Figure 8, maximum gradient at 50 G/cm, 17 cm in diameter, slew rate = 2100 mT/m/s) [43] to allow high spatial resolution imaging for mice. This high performance insert gradient set was used for phantom and mouse imaging [27]. A quadrature birdcage RF coil for mice (Figure 9, 3 cm in diameter and 6 cm in length, Morris Instruments, Ottawa, ON) was used for 3He imaging (97.3 MHz) [27]. The coil was driven using a 8 kW AMT 3 T90 RF power amplifier (GE Healthcare, Waukesha, WI) using a manufacturer-supplied T/R switch and preamplifier tuned to 97.3 MHz. Typical rectangular 90° RF pulse lengths were 100 μs [27]. A VFA RF pulse trajectory was employed to reduce blurring due to RF de-polarization [47, 52]. Center-out sampling in the phase-encoding direction was used [52]. The VFA pulses were calibrated as described in [48].

### 3.5 Image reconstruction

A custom-made IDL6.4 (ITT Visual Information Solutions, Boulder, CO) routine was used for off-line Fast Fourier transformation (FFT) of the k-space data. For reconstruction of the X-Centric data, a combination of both halves of k-space was applied prior to the inverse Fourier transformation as follows. To form a full line of k-space data in the read-out direction, the two half-echo data sets were combined [17]. Each half-echo data set included two additional points (BWread = 62.5 kHz dwell time was 8 μs) prior to kx = 0 which can be thought of as partial-echo factor of 0.505 for each half.

Acquisition of a few extra kx data points is helpful since the presence of background gradients can shift the origin of k-space by one or more points resulting in mis-sampling or causing image artifacts [57]. These extra k-space points (only two) were removed before combining the two halves of k-space prior to reconstruction [17]. The disadvantage of this approach is that this leads to a b-value three times larger and longer TR due to the extra sampled points. However, phantom results demonstrate that this increased b-value (b = 0.061 s/cm2) is still quite practical, ensuring minimal mis-sampling even in the presence of significant background-gradients and gradient-induced eddy currents [17, 27, 39]. Collection of the extra four kx points (two points per half) for each kx line results in a 1.5% over-sampling factor for a 256 × 256 matrix size as an example [27].

### 3.6 In-vivo imaging

X-Centric sequence parameters for the highest image resolution (78 μm) are shown in Table 2. In-vivohigh-resolution 2D whole lung projection 3He mouse lung images obtained for partial-echo (top panels) and X-Centric (bottom panels) FGRE acquisitions for three different matrix sizes are shown in Figure 10 (axial view) and Figure 11 (coronal view). Images were obtained during a 1 s breath-hold following three wash-out 3He pre-breaths. It is easy to see that both the axial and coronal images of the mouse lung obtained at high spatial resolution (78 μm) with X-Centric demonstrates visible signal in the major airways. Axial and coronal images of mouse lung obtained for the same resolution setting with partial-echo FGRE show no signal in the trachea as well as lower SNR in the lung periphery. The mean SNR for X-Centric images computed from the mouse lung airways and parenchyma [58] were 25 and 14 respectively [27]. The corresponding mean SNR for the partial-echo FGRE computed for airways and parenchyma were 1.8 (which is less than the Rose criteria of SNR = 5, [59]) and 8 respectively [27]. The large b-value of the partial-echo FGRE led to complete signal destruction in the major airways where 3He experiences near-free diffusion (D = 2 cm2/s). Besides this, a relatively long TE (1.2 ms [27]) in combination with a shorter T2* (Table 1) in the mouse lung trachea also significantly contributes to signal decay together with diffusion losses. In comparison to the larger airways, the lung parenchyma within the images obtained with partial-echo FGRE still shows some signal, which is not surprising considering that the 3He diffusion was significantly restricted here (D = 0.21 cm2/s [37]) and T2* value was long enough (Table 1).

The larger SNR in the major airway (X-Centric mouse lung images) compared to the lung parenchyma is likely due to the higher 3He density in trachea and reasonably short TE used in X-Centric sequence (0.7 ms [27]).

It is useful to calculate the SNR efficiency (ϒ) of the X-Centric and partial-echo FGRE acquisitions was using the following equation [60]:

ϒ=SNR/ΔxΔyTscanE14

where Tscan is the total imaging time (in s) and SNR was measured based on the mean of Region of Interests (ROI) placed in the rodent lung airways and parenchyma divided by the standard deviation measured in the background (noise) [58]. Thus, X-Centric was 13.4 times more efficient than the partial-echo FGRE sequence for the case of the major airways [27] and 1.7 times more efficient in case of MR imaging of parenchyma [27] (Figure 11).

### 3.7 Limitations

One of the limitations of the X-Centric method is the need to use the FFT for image reconstruction from two half-echo k-space samples rather than a single k-space. Instead, one can use a single half-echo k-space and Partial Fourier Reconstruction (PFR) [61]. To test this idea we used the Projection Onto Convex Sets (POCS) method [62] to reconstruct a phantom image for 130 × 256 (2 + 128 × 256) matrix size (Figure 5b). Figure 12 shows phantom images obtained with PFR (left column) and FFT (right column) for three different matrix sizes. The following SNR values were obtained for the presented images: SNR = 39 (160 × 256, PFR); SNR = 54 (160 × 256, zero-filling in x-direction and then FFT); SNR = 45 (130 × 256, PFR); and SNR = 58 (256 × 160, zero-filling in y-direction and then FFT). The results confirm that the FFT with zero-filling approach gives better SNR than the PFR method for 160 × 256 matrix (39 vs. 54). We have calculated the efficiency of half-echo acquisition (130 × 256, SNR = 45) reconstructed with PRF and X-Centric (256 × 160, 62.5% under-sampling in y-direction, SNR = 58) reconstructed with FFT in order to compare imaging methods. Thus, the calculated efficiency ratio is 0.87 (Eq. (14), with Tscan = 256 and 160 + 160 y-gradient steps for the half-echo and the X-Centric respectively). This shows that a half-echo acquisition is 13% less efficient than X-Centric, which reflects smaller SNR due to the 50.1% sampling vs. 62.5% sampling for the case of X-Centric, but half-echo acquisition is ∼200 ms faster due to the doubling of y-gradient steps for X-Centric. We believe that our half-echo acquisition vs. X-Centric results indicate that the half-echo approach is 20% faster but 10% less SNR efficient than X-Centric. A structured phantom needs to be used for probing the potential smoothing effects and resolution implications following the Partial Fourier Reconstruction.

### 3.8 Future role of X-Centric

The focus of this chapter is a presentation of a fast and SNR efficient imaging method for high spatial resolution imaging of mouse lungs that include both the airways and lung parenchyma. Such a technique can be potentially used for mapping the morphological changes and ventilation heterogeneity associated with rodent models of asthma or ovalbumin-challenged model (OVA). Figure 13 shows an example of ventilation heterogeneity maps obtained for a sham mouse (right column) and OVA treated mouse (left column). The figure compares increased (a and c) and decreased (b and d) ventilation heterogeneity indexes from a sham mouse (left column) and an OVA mouse (right column) under similar increases in airways resistance (104% vs. 93%, respectively) and elastance (21% vs. 27%, respectively). The white and red lines in the figure outline the lung contour before and after the Methacholine inhalation (MCh) challenge, respectively. Airway resistance and tissue elastance were measured immediately following each scan. Lung images were then analyzed using local standard deviation (SD) in sliding (9 × 9 pixels2) ROIs. The ventilation heterogeneity index (VHI) was given by:

VHI%=SDpostMChSDpreMChSDpreMChx100%E15

The post-MCh lung function data were normalized by the pre-MCh values to give the percent change. The sham mouse and an OVA mouse data are shown only as an example of the potential X-Centric pre-clinical application. However, data suggested that ventilation heterogeneity following MCh challenge may be commensurate with traditional lung function testing to access the airway hyper-responsiveness in the OVA mouse model. The observed heterogeneity in ventilation distribution could potentially provide a novel endpoint to study disease modification in asthma, as well as for better diagnosis and classification of asthma in the clinic.

The X-Centric method described here should also be beneficial for clinical hyperpolarized lung MRI and other applications where short-TE techniques are needed. Table 2 helps to estimate the acquisition time for BWread = 62.5 kHz and FOV = 200 × 200 mm2 (pediatric study [63]) for two matrix sizes frequently used in human studies [6]: 128 × 128 × 14 slices and 256 × 256 × 14 slices (3D case) yielding resolutions of 3.1 × 3.1 × 15 mm3 and 1.6 × 1.6 × 15 mm3 at FOV = 400 × 400 × 15mm3. With the assumption of partial sampling in the phase-encoding direction and taking into account that the maximum strength of the clinical gradient system is typically 4–5 G/cm with a maximum slew rate of 200 T/m/s [6], one can calculate the acquisition to be around 5 s for a 128 × 128 matrix and 16.7 s for a 256 × 256 matrix. As the 256 × 256 matrix at a FOV of 20 × 20 cm2 corresponds to 781 μm nominal resolution, this should lead to 70% signal decay in airways due to diffusion-weighting according to [22] for the case of full-echo FGRE. The X-Centric approach could eliminate 90% of the signal loss due to incidental diffusion-weighting in the major airways. Parallel imaging approaches [64] or compressed sensing approaches [65, 66] can also speed up data acquisition even more and are a straight-forward extension of the X-Centric approach described here.

Until now, X-Centric has not been used in any pediatric studies; nevertheless, hyperpolarized gas MRI is gradually becoming an important research tool, allowing for longitudinal observation of lung diseases in children such as Cystic Fibrosis (CF). Figure 14 shows 3He MRI lung images obtained from a 17-month old patient with CF. Note, the regions of hypo-intensity are difficult to see at this resolution, but using X-Centric, one should be able to gain diagnostic value and better enhance the contrast in these regions. We believe that X-Centric can be very useful for not just CF but the longitudinal observation of other lung diseases like Bronchopulmonary Dysplasia (BPD). Abnormal prematurity at birth (<28 weeks gestation) frequently requires neonatal intensive care and mechanical ventilation due to respiratory distress syndrome due to limited or lack of lung surfactant and incomplete lung development. Ventilator-associated lung injury, along with the extreme prematurity of the lungs themselves, causes abnormal and irregular lung growth (BPD) [67]. The relatively short signal life-time caused by the need to scan newborns inside their neonatal incubators (causing B0 field distortions, especially at high field) and also the need for smaller FOV make the X-Centric approach ideally suitable for high-resolution neonatal lung MRI.

## 4. Use of X-centric with other MR-visible nuclei

### 4.1 Gas phase 129Xe MRI

3He is a rare and expensive isotope (Table 1), so a worldwide clinical translation of the 3He MRI method is quite questionable, especially when there is a relatively inexpensive alternative—129Xe MRI. In contrast to the 3He isotope, the 129Xe isotope is much more abundant (∼26% natural abundance) and cost efficient (Table 1). In addition, recent improvements in the xenon polarization process have enabled sufficient increase in the level of 129Xe polarization (>30%) and production volume of the polarized gas (∼2 L per hour) [68, 69]. Presently, 129Xe lung MRI is translating towards a clinical tool, and it has recently been used as a clinical tool (along with 3He MRI) in the United Kingdom [70]. The North-American xenon consortium [71] expects 129Xe MRI to be the Food and Drug Administration approved in the first quarter of 2020.

Presently, the main directions of 3He MRI are pediatric and neonatal studies (using relatively small doses of 3He) mainly due to the cost of the 3He isotope and the high quality of the 3He lung images. In turn, 129Xe MRI is much more suited for adult lung imaging; however, the low SNR of 129Xe images (compared to 3He) impede the development of many novel acquisition schemes that are highly sensitive to SNR [72], such as isotropic voxel static-ventilation imaging and accelerated multiple b-value diffusion-weighting MRI [73]. Mentioned above, recent improvement in xenon polarization [68, 69] provides a way to develop advanced imaging methods previously not possible. This is especially critical for investigations in patients with asthma and COPD. Here we report the isotropic-voxel, high-resolution 3D 129Xe static-ventilation image acquired in a single 16 s breath-hold, with the necessary and sufficient SNR to quantify ventilation defect percent [74] estimates (an efficient asthma and COPD biomarker). To the best of our knowledge, a voxel size of 3 × 3 × 3 mm3 is the highest currently achieved resolution for 129Xe human lung images. Figures 15 and 16 show the axial and coronal lung images obtained for a healthy volunteer using an isotropic voxel 3D FGRE pulse sequence. The central slice SNR was around 30 for both axial and coronal view images. The gas phase of 129Xe has a relatively low free diffusion coefficient (0.14 cm2/s in air, Table 1) and relatively long signal life-time (10–30 ms depends on the field strength [75]), so the benefits of the X-Centric approach (overcoming a high diffusivity and short T2*), may not be very obvious for the case of healthy subjects compared to the traditional FGRE sequence. However, lungs of patients with severe emphysema, COPD (smokers or ex-smokers), BPD and Alpha-1 Antitrypsin Deficiency may show regions with significant terminal airways destruction, changing the diffusion regime to nearly-free diffusion (xenon D ≈ 0.22 cm2/s when mixed 50/50 with helium-4 [6]) in such regions, naturally leading to substantial local reduction of T2* (up to 50% reduction due to the fast motion through the local Bo inhomogeneities within the lung airspaces) [73] especially at high field. X-Centric, is a short-TE method that should be able to better visualize lung regions with short T2*; and therefore, a better tool for helping to make clinical decisions and for therapy planning.

### 4.2 Dissolved phase 129Xe MRI

Unlike its counterpart 3He, 129Xe is highly soluble in a variety of solvents and biological materials, and exhibits a large range of chemical shifts within these distinct chemical environments (Ostwald solubility coefficients of 0.17 [40]). Of particular interest is dissolved phase 129Xe residing in and exchanging with lung tissue (tissue barrier) and red blood cells (RBC) [76, 77, 78]. The exchange can be measured using MRI and it is also possible distinctly resolve and quantify 129Xe within the tissue barrier and RBC. Exploiting these signals for diagnostic imaging purposes can be difficult since for example, only about 2% of inhaled xenon dissolves into the tissue barrier and RBC [76, 79], and the corresponding T2* values are often short (∼2 ms) [80]. There is an urgent need for efficient short-TE methods to acquire these short T2* signals and enable deeper study of a number of lung abnormalities such as ventilation/perfusion mismatch [81, 82] and of other organs including xenon brain perfusion imaging [72, 83] and imaging of the xenon-encapsulated Cryptophane-A functionalized with anti-cancer drugs [84].

The feasibility of the X-centric method was initially demonstrated using 3He for high-resolution rat lung imaging [39] indicating its ability to overcome diffusion induced signal attenuation within the trachea and substantially reducing TE and td as well as the magnitude of the imaging gradient [27]. The low diffusivity and low gyromagnetic ratio of 129Xe (Table 1) suggest that diffusion decay due to the imaging gradients would not be an obstacle for high-resolution 129Xe gas MRI; however, the relatively short T2* of the dissolved xenon remains a problem, as it leads to signal decay and contributes to image blurring simultaneously. Clearly, short-TE pulse sequences such as spiral [85, 86] or X-Centric [17, 27] are absolutely necessary to acquire images of xenon dissolved in lung tissue and blood.

Figure 17b shows dissolved phase hyperpolarized xenon obtained from within a water-filled resolution phantom (Figure 17a) on a clinical 3 T MRI scanner (maximum gradient strength was 5 G/cm, sequence parameters are shown in Table 2). The 129Xe image shows that X-Centric can be used to image dissolved phase xenon efficiently (within a breath-hold) and with reasonable signal-to-noise ratio (SNR = 12). After reconstruction, the smallest geometric features of the resolution phantom in the X-Centric image are discernable (yellow arrows). It should be noted, that the available SNR achievable when imaging dissolved phase 129Xe principally depends on concentration, polarization and relaxation times and even physiological parameters such as blood flow and blood oxygenation. The quality of the polarizer, solubility of the solvents used and chemical environments all play a role. Nevertheless, techniques employing significant reductions in echo-times can deliver higher SNR than is typically achievable for spins within environments where their transverse relaxation times are significantly shortened. Though, the phantom dissolved-phase spins had estimated T2*-values of roughly 20 ms, the potential for using X-Centric for imaging xenon within the tissue barrier, RBC and brain (T2* < 2 ms) is high and the prospects for high-resolution dissolved phase imaging (currently not available) are demonstrated.

### 4.3 Static-ventilation 19F MRI

The limited use of the hyperpolarized 3He/129Xe MRI imaging modality can be partially explained by its need for utilizing expensive isotopes and polarizers, the latter requiring specially-trained personnel. Inert inhaled fluorinated gas MRI can be a promising and less expensive (Table 1) alternative to hyperpolarized 3He/129Xe MRI. 19F MRI does not require the use of rare isotopes (100% naturally abundant, Table 1) or expensive polarizers. 19F gases (such as perfluoropropane (C3F8) and sulfur hexafluoride (SF6) [17, 18]) can be imaged using high field (≥1.5 T) clinical MRI scanners; however, a multi-nuclear amplifier and receiver along with RF coils tuned to the 19F frequency are still required [18, 19]. Fluorine-19 has a large gyromagnetic ratio (∼95% of 1H, Table 1), both SF6 and C3F8 have several 19F atoms per molecule, and have relatively short T1 values making these highly attractive options for researchers (Table 1, [87]). Short T1 translates to overcoming their low thermal polarization [17, 88] values by rapid signal averaging. Additionally, because it is not hyperpolarized, Fluorine-19 gases can be mixed with O2 without polarization losses, preserving the image quality patient safety simultaneously during a breath-hold scan [89]. Inert fluorinated gases have very short T2* values (T2* < 2 ms [87]), therefore imaging methods with short echo times are preferable. As we demonstrated previously, the X-Centric acquisition can be reliably used for phantom and in-vivomeasurements using both 3He and 129Xe gases. Similar to the dissolved 129Xe case, a short apparent transverse relaxation time specific to the fluorinated gas can potentially lead to fast signal attenuation and image blurring. As X-Centric is a short-TE sequence it should be able to ensure sufficient SNR of the fluorinated gas images, and also minimal image blurring. Figure 18a and b show an SF6 high-resolution phantom (Figure 14a) and rat lung images obtained using X-Centric on a clinical 3 T MRI scanner (maximum gradient strength was 5 G/cm, in-vivoimage sequence parameters are shown in Table 2). In-vivodata acquisition was synchronized to breath-hold durations using the ventilator [17]. Rat lungs were saturated with an SF6/O2 80/20 gas mixture during 3 min of normal free breathing at a breathing rate of 60 breaths/min. After 3 min, the fluorinated gas/O2 mixture was suspended followed by a 10 sec breath-hold to obtain in-vivorat lung image (Figure 18b) [17]. Inert fluorinated gas images confirm that X-Centric ensures sufficient SNR (14 and 40 for the phantom and in-vivoimages respectively) and resolution (phantom images shows all details) of fluorinated gas images, when apparent transverse relaxation time is less than 2 ms. These results suggest the possibility for translating these techniques directly to human use because they are demonstrated using a clinical scanner using clinical gradient coils.

### 4.4 Dynamic 19F MRI

19F gas MRI also has a high potential for successful dynamic lung imaging [88, 90] due to the fact that fluorinated gases can be mixed with O2 which helps to restore the initial magnetization faster (decreasing T1), in contrast to the hyperpolarized gases. Furthermore, imaging times are shortened and breath-holds are more easily tolerated by the patients [89]. A technique like this permits the acquisition of regional fractional-ventilation [23, 48, 91] which can be very useful as a radiation free alternative to CT for detecting gas trapping in lung diseases such as lung fibrosis and COPD [88]. Recently, free-breathing 19F (C3F8) dynamic lung imaging has been demonstrated in human lungs [89]. Figure 19 (top panel) shows the gradual wash-out 19F gas within the 19F MRI lung images obtained from a COPD patient for eight wash-out breaths [89]. As expected, each new wash-out breath of air replaced some volume of the fluorinated gas in lung, so the signal intensity of the resulting images was gradually attenuated. The following equation can be fitted to this wash-out data [48]:

Sn=S01rnE16

where S0 is the initial signal, nis the breath number, S(n) is the signal intensity after a certain number of wash-out breaths and ris the fractional-ventilation parameter [17, 48]. In turn, rcan be expressed as the fraction between new 19F incoming to lung, and the total volume of 19F (Vtotal) [17, 48]:

r=VnewVtotal=VnewVold+VnewE17

where Vnewand Voldare the “new” and “old” gas volumes in each voxel at each breath. In general, Eq. (16) can be approximated with the stretched exponential (or Kohlrausch) function [92]:

Sn=expnrβE18

where 0 < β ≤ 1, and ris the fractional-ventilation. This simple function has been recently used to combine the CS method with under-sampling in the parametric direction [93], which permitted an acceleration factor (AF) of 7 in multi b-value diffusion-weighted 3He MRI measurements [73]. Generally, a reasonable choice of the AF value depends on a number of the images in parametric direction. The main idea behind this relationship is to keep the reconstruction errors low [93]. As we stated above in this chapter, X-Centric is Cartesian pulse sequence, which can be naturally used for under-sampling in imaging (phase-encoding) and parametric (time) directions following CS with the stretched exponential method [93]. To our knowledge this is first record of the possibility of using CS combined with the stretched exponential method for dynamic lung imaging studies. Figure 19 (bottom panel) shows an under-sampled X-Centric k-space pattern ensuring AF = 14. Remarkably, this means that only 8% of k-space, or five k-space lines (out of 64) are sampled to achieve a “full dataset” and each image in time can be acquired 13 times faster, which obviously helps to avoid the motion artifacts common to free-breathing data acquisition. Additionally, X-Centric can potentially ensure a much shorter TE compared to the spoiled gradient-echo sequence (TE = 1 ms) used to obtained 19F human lung images [89]. Retrospective X-Centric-like under-sampling of the 19F dynamic rat lung in-vivodata (64 × 64 matrix size, nine wash-out images [17]) available to us has confirmed that AF = 14 led to less than a 12% reconstruction difference (pixel-by-pixel comparison between the fully-sampled image and one reconstructed from the retrospectively under-sampled k-space). Therefore, the choice of X-Centric for dynamic 19F gas MRI lung imaging will speed up imaging time and improve image quality making it an optimal choice for this kind of imaging.

The remaining obstacle preventing wide-spread high-resolution 19F MRI with or without X-Centric is the generally low SNR of 19F lung images. In this particular case, the further improvement of the image quality strongly depends on a hardware component, specifically from RF coil systems. It has been recently demonstrated that the use of a multi-channel phased-receive array can significantly improve the image quality of the 19F MRI lung images [94]. This is a very promising result suggesting that the combination of the X-Centric method with advanced array-type RF coil system will permit future high-resolution 19F MRI human lung imaging in both static-ventilation and dynamic studies.

## 5. Conclusion

In summary, hyperpolarized 3He X-Centric MR imaging with high spatial resolution was proven to be a robust technique in phantom and mouse lung measurements yielding a nominal resolution of 39 μm and 78 μm respectively. In particular, mouse major airways with less restricted diffusion of 3He (D∼ 0.9 cm2/s) could only be visualized with the X-Centric method. Note, that the high nominal resolution (78 μm) in in-vivoscans was achieved within 1 s breath-hold. These results suggest that the X-Centric method can potentially fill a gap and need for high spatial and temporal resolution imaging method for the small model of asthma studies.

Another beneficial feature of the X-Centric sequence is the significantly reduced echo-time, so the method can be used for imaging of nuclei with short signal life-times such as hyperpolarized 129Xe dissolved within the lung tissue barrier and/or Red Blood Cells, and inert fluorinated gas (SF6) inside the lungs. In both cases the T2* value is less than 2 ms. The feasibility and future potential of short-TE X-Centric-based X-Nuclei measurements were demonstrated using 129Xe dissolved in water as well as SF6 within rat lungs.

## Acknowledgments

The authors would like to thank the following individuals for assistance with MRI experiments and data analysis: Grace E. Parraga, Giles E. Santyr, Mitchell S. Albert, Marcus J. Couch, Tao Li and Iain K. Ball. A special thanks to Ben T. Chen for providing support for the hyperpolarized 3He gas, animals, and ventilator.

## Conflict of interest

The authors do not have any conflict of interest.

## Notes/thanks/other declarations

We thank Michael Völker for providing the MatLab code (Projection onto Convex Sets (POCS), 2013) for image reconstruction. We thank Abascal et al. for providing the MatLab code (Signal Decay Into the Reconstruction (SIDER), 2017) for image reconstruction.

## Nomenclature

AF

acceleration factor

bandwidth

BALB/c

laboratory-bred strain of the house mouse

BPD

bronchopulmonary dysplasia

C3F8

perfluoropropane

CF

cystic fibrosis

CFA

constant flip angle

COPD

chronic obstructive pulmonary disease

D

self-diffusion coefficient

19F

fluorine-19

FFT

fast Fourier transformation

FGRE

FOV

field of view

FWHM

full width at half maximum

GE

general electric

1H

hydrogen-1

3He

helium-3

MCh

methacholine inhalation challenge

MRI

magnetic resonance imaging

O2

oxygen

OVA

ovalbumin-challenged model

PFR

partial Fourier reconstruction

POCS

projection onto convex sets

RBC

red blood cells

ROI

region of interest

RF

SD

standard deviation

SDpost-MCh

post-methacholine challenge SD

SDpre-MCh

pre-methacholine challenge SD

SF6

hexafluoride

SNR

signal to noise ratio

SNRnor

normalized signal to noise ratio

SNRtheor

theoretical signal to noise ratio

T1

longitudinal relaxation time

T1,0

T1 expected in the absence of O2

T2*

apparent transverse relaxation time

Tscan

total scan (imaging) time

TE

echo time

T/R

TR

repetition time

VFA

variable flip angle

VHI

ventilation heterogeneity index

129Xe

xenon-129

chapter PDF
Citations in RIS format
Citations in bibtex format

## More

© 2019 The Author(s). Licensee IntechOpen. This chapter is distributed under the terms of the Creative Commons Attribution 3.0 License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

## How to cite and reference

### Cite this chapter Copy to clipboard

Matthew S. Fox and Alexei V. Ouriadov (March 6th 2019). High Resolution <sup>3</sup>He Pulmonary MRI, Magnetic Resonance Imaging, Lachezar Manchev, IntechOpen, DOI: 10.5772/intechopen.84756. Available from:

### Related Content

#### Magnetic Resonance Imaging

Edited by Lachezar Manchev

Next chapter