Introduction
Molecules with incorporated Fluorine-19 (
19F) receive much interest as Magnetic Resonance Imaging (MRI) contrast agents in heteronuclear or hot-spot imaging [
1‐
3]. With
19F MRI, background-free
in vivo images can be acquired, which can be combined with conventional proton MRI for precise anatomical localization. Moreover, the
19F signal is unambiguous; the signal measured is linear to the concentration of the imaging agent [
4], and allows for quantification from the image data [
5]. This differentiates heteronuclear contrast agents from conventional contrast agents, such as gadolinium chelates and super paramagnetic iron oxide particles, which create contrast by altering the existing signal [
6,
7].
19F MRI does suffer from sensitivity issues, limited number of clinically approved imaging agents and the need for dedicated hardware. Currently,
19F MRI has been successfully implemented in preclinical studies in which it is used for cell tracking and inflammation imaging [
8‐
12]. Simultaneously, clinical hurdles are being overcome with the first in-human studies and good manufacturing process-certified production methods [
13,
14].
In nearly all preclinical studies, the animals need to be anesthetized during imaging to prevent unwanted movement and discomfort. This is most frequently done with isoflurane (ISO) inhalation anesthesia. As the name implies, ISO is a fluorinated anesthetic agent and at therapeutic levels its concentration is sufficient to be observed in MRI [
15]. Therefore, the fluorine atoms in ISO are a problem for
19F MRI as conventional MRI-sequences cannot distinguish ISO from the fluorinated contrast agent leading to artifacts in imaging. The ISO artifact signal can lead to misinterpretation of the images. Furthermore, when imaging agent and ISO signals overlap, quantification is no longer accurate.
The simplest way of avoiding ISO artifacts in
19F MRI is to use a different anesthetic agent without fluorine atoms (e.g. ketamine, pentobarbital). However, ISO is very easy to use, widely available and has relatively little influence on cardiovascular function and cerebrovascular blood flow [
16,
17]. Therefore, ISO is preferred above all other anesthetic agents, even in
19F MRI applications.
To be able to use ISO in 19F MRI applications, we present three methods that can be used to avoid imaging artifacts from ISO and can be readily implemented on most MR systems. The ISO signal is either shifted out of the image by using a narrow acquisition bandwidth, suppressed by using a frequency selective suppression pulse or not excited by using a narrow excitation bandwidth in a 3D acquisition. These methods are tested in mice anaesthetized with ISO and injected with perfluoro-15-crown-5 ether (PFCE) containing nanoparticles (NPs) as the 19F signal of interest. The methods either shift the ISO signal away from the region of interest, suppress the ISO signal or avoid exciting the ISO altogether.
Theory
In image formation in MRI, the MR signal can be spatially localized during excitation with slice or slab selection and during signal reception by a combination of phase and frequency encoding. Whenever magnetic field gradients are used for slice-selective excitation or for spatial encoding during signal reception, chemical shift displacement artifacts (CSDA) occur when signals are present at multiple resonance frequencies. During slices selection, the signal of interest, at the scanner’s reference or carrier frequency, is excited or refocused at the slice position of choice. Off-resonant signals (ISO in our case) experience this excitation or refocusing at a displaced slice position, of which the distance from the intended position depends on the frequency difference from the off-resonant signal to the carrier frequency and the bandwidth of the selective radiofrequency pulse. During signal reception, the signal of interest is projected in its real location in the image, but all off-resonance signals experience a chemical-shift dependent displacement in the image along the frequency encoding direction. With its wide chemical shift range of resonances, different signals in
19F MRI can be largely displaced relative to each other in excitation and reception. In our
19F MRI example, three resonance groups are present (PFCE singlet at − 91.5 ppm, ISO CF
3-group singlet at − 82.9 ppm and J-coupled ISO CF
2-group of peaks at − 89.9 ppm, Fig.
2A). When a PFCE and ISO phantom is imaged with a standard pulse sequence with the carrier frequency at − 91.5 ppm, the acquired image contains a PFCE signal in the real location and a displaced signal from the CF
3-group of ISO (Fig.
2B). Note that when imaging
in vivo, both the CF
3 and CF
2 groups from ISO are visible and can be present at multiple unexpected locations within the animal. Since the ISO signals have a higher resonance frequency compared to PFCE and, in this example, the frequency encoding direction is from bottom to top, the ISO signals shift upwards. The degree of shift, expressed in pixels, is the direct result of the frequency encoding acquisition bandwidth and the frequency difference between ISO signals and the imaging reference frequency (i.e. the PFCE frequency) and can therefore be defined by:
$$pixel\;shift=\frac{fISO-fPFCE(Hz)}{\textit{acquisition}\;BW(Hz/pixel)}$$
Similarly, if slice or slab selection is used, and the imaging reference frequency is the PFCE frequency, the CSDA causes a relative shift in slice excitation position of ISO, which depends on the bandwidth of the excitation radiofrequency pulse of:
$$\textit{relative}\;slice\;\textit{position}\;shift=\frac{fISO-fPFCE(Hz)}{\textit{excitation}\;BW(Hz)}\ast100\%$$
In this work, we use the known frequency differences between a PFCE-containing imaging agent and ISO signals (i.e., CF3 offset 4042 Hz and CF2 offset 2820 Hz from PFCE at 11.7 T), and the predictable CSDA, to avoid ISO interference.
Discussion
19F MRI always involves the introduction of an imaging agent, as there is no visible fluorine present in biological systems. Non-ambiguous and quantitative imaging is the goal with
19F MRI; however, unwanted fluorine signals coming from ISO are often a complicating factor. Various fluorinated molecules are being used as imaging agents for
19F MRI. For some fluorinated molecules, it can be expected that ISO CSDA is not an issue because of the large frequency difference between these molecules and ISO (e.g., perfluorocyclohexane and oxycyte [
21,
22]). However, the most commonly used agents (e.g., PFPE, PFOB, PFCE, PERFECTA) are expected to suffer from interfering ISO signals due to their resonance frequencies which are close to that of ISO [
18,
23].
Currently, most studies using
in vivo 19F MRI do not go into detail on whether and how ISO CSDA are avoided. Often insufficient measures are taken to avoid ISO artifacts: (1) Scanning early in the anesthesia period, with less ISO accumulation and therefore slightly lower signal. (2) High imaging agent signal. (3) Different visibility of ISO through different relaxation times at different magnetic field strengths [
24]. Regardless, ISO signal is present and could therefore hamper image interpretation and quantification.
In literature, a number of approaches have been used to avoid ISO artifacts. However, implementation of these strategies might not be straightforward or additional post-processing is needed. For example, chemical shift encoding approaches use a multi-echo acquisition and a modeled image reconstruction to remove CSDA from the image [
25,
26]. This approach can reduce SNR in single resonance PFCs and needs implementation of a novel sequence and post-processing. Deconvolution methods use an algorithm-based process to correct for the distortion caused by the known CSDA in post-processing [
27,
28]. This technique, however, suffers from blurring, increased noise in low SNR images and artifacts that could impede image interpretation and quantification, especially in the complex and broad CSDA encountered with ISO. Finally, some studies have used injection anesthesia as an alternative to ISO, consequently missing out on the key benefits of using ISO compared to injection anesthesia [
8,
29‐
31].
In this work, three distinct strategies to avoid ISO CSDA are described, all of which can be applied in
19F MRI without a special sequence or post-processing (Table
2). These methods rely on the differences in resonance frequencies between ISO and the imaging agent. The illustrated explanations in this work help understand the ratio behind these strategies. Overall, these strategies will image only one resonance frequency resulting in SNR loss in multiple-resonance PFCs such as PFOB. The methods need to be adapted to the user’s magnetic field strength, fluorine molecule, object size and resolution depending on the specific situation of the user [
32]. The influence of magnetic field strength on the applicability of the strategies to avoid ISO artifacts is not investigated here. At lower field strength, the frequency difference between the signal of interest and the ISO signal will be smaller. As result, the bandwidth needed to achieve the same effect also needs to be smaller. This will be challenging for the shift out of plane sequence, but is very well feasible for the two other strategies.
Table 2.
Pros and cons of the three ISO-avoidance strategies
Shift out of plane | 2D sequence allows for short acquisition times (when signal allows this) | Signal loss due to long TE |
No need for new sequences and/or post-processing | Increase in TR due to longer TE |
Very easy implementation | Blurring due to acquisition bandwidth per pixel < peak width |
| Bigger objects need more ISO shift which aggravates these downsides |
Suppression pulse | 2D sequence allows for short acquisition times (when signal allows this) | High SAR |
Freedom in excitation and acquisition bandwidths, short TE | Slight increase in minimal repetition time |
No need for new sequences and/or post-processing | Sub-optimal shimming might result in incomplete suppression |
Easy to implement | B1-inhomogeneties can lead to a non-90° suppression pulse which can result in incomplete suppression |
Do not excite | Increase in SNR due to 3D approach | Phase encoding in two directions results in movement artifacts in two directions |
No distortion, short TE/TR | Phase encoding in two directions can result in fold-in artifacts in two directions |
No need for new sequences and/or post-processing | |
Easy to implement | |
Compared to the phantom NMR spectrum of Fig.
2A,
in vivo line broadening of PFCE is observed (Fig.
2C). This line broadening can be the result of inferior shimming and/or PFCE molecular mobility differences
in vivo compared to a simple phantom tube. For ISO, not only line broadening, but also a different
in vivo chemical shift is observed (Fig.
2A,
C). This is the result of binding and dissolving of ISO in lipids, changing its molecular mobility and chemical environment [
24].
Shifting the ISO signal out of the image, by using a narrow acquisition bandwidth, successfully removes the ISO artifact from the image. This approach has downsides. Signal might be lost due to a longer TE, and increased blur may occur due to signal decay during the frequency readout. Signal loss is in part negated by lower noise due to the low acquisition bandwidth, but blurring of the image occurs because the acquisition bandwidth per pixel is smaller than the peak width of PFCE
in vivo. These artifacts are clearly visible in Fig.
3B, where the PFCE signal can be seen projecting outside of the animal.
Suppressing the ISO signal by using a frequency selective suppression pulse resulted in complete ISO signal suppression. If this approach fails, possible reasons could be the quality of the shim or B1 field inhomogeneities greater than the chemical shift differences. It might be possible to improve the robustness of the suppression pulse approach by using a suppression pulse flip angle slightly larger than 90° to compensate for the recovery of signal between the suppression pulse and excitation pulse (in our case 3 ms). Moreover, the addition of the suppression pulse leads to an increase in SAR which can result in heating of the subject.
The do-not-excite sequence, using a 3D acquisition scheme has two major advantages. First, an increase in SNR (151 ± 28%) due to the 3D approach. Second, applicability at lower magnetic field strength as the excitation bandwidth can easily be further decreased. In the do-not-excite-ISO sequence, the aim is to use the slab-selective magnetic field gradient to temporarily have the off-resonant ISO signal resonate at frequencies out of reach with the applied RF pulse outside of the coil with an appropriate combination of pulse bandwidth and chemical shift difference, as illustrated in Fig.
5D. When the ISO signal resonance frequencies are moved outside of the RF coil sensitivity profile, they will not be excited and cannot produce a signal during acquisition. By defining a relative — rather than absolute — shift in slice or slab position as we do in Eq. (2), the shift description is independent of the applied magnetic field strength, but expressed in a factor or percentage relative to the intended slice or slab thickness. With known RF pulse bandwidth, the applied magnetic field gradient defines the slice or slab thickness, and the relative shift defines the position of the off-resonant excitation. In our example, a 32-mm slab is used. With a 1-kHz excitation bandwidth and a chemical shift difference of 2820 Hz, this means that the center of the excitation slab at the ISO-frequency is located 2.82 slabs, or 90 mm, away from the location of the on-resonance PFCE. With a 40-mm coil length ISO, signal frequencies are beyond the coil’s excitation profile. This method has been used successfully previously [
19,
33,
34]. If a coil setup is used in which transmit and receive coils are separated, usually the transmit RF coil is larger than the local receive coil. In such cases ISO signals can still be excited, but perhaps not received with the local receive coil.
In summary, we recommend to report on the method used to avoid ISO CSDA in all 19F studies that use ISO as anesthetic. In order to understand how ISO CSDA are avoided, the method section should include the following parameters: Frequency difference between the fluorinated molecule of interest and ISO, excitation and acquisition bandwidth and pulse shape, 2D or 3D sequence and coil length. When using a suppression pulse the bandwidth, offset, flip angle and shape of this pulse should be provided.
Publisher's note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.