Four-Channel Surface Coil Array for 300-MHz Pulsed EPR Imaging: Proof-of-Concept Experiments

Abstract

Time-domain electron paramagnetic resonance imaging is currently a useful preclinical molecular imaging modality in experimental animals such as mice and is capable of quantitatively mapping hypoxia in tumor implants. The microseconds range relaxation times (T₁ and T₂) of paramagnetic tracers and the large bandwidths (tens of MHz) to be excited by electron paramagnetic resonance pulses for spatial encoding makes imaging of large objects a challenging task. The possibility of using multiple array coils to permit studies on large sized object is the purpose of the present work. Toward this end, the use of planar array coils in different configurations to image larger objects than cannot be fully covered by a single resonator element is explored. Multiple circular surface coils, which are arranged in a plane or at suitable angles mimicking a volume resonator, are used in imaging a phantom and a tumor-bearing mouse leg. The image was formed by combining the images collected from the individual coils with suitable scaling. The results support such a possibility. By multiplexing or interleaving the measurements from each element of such array resonators, one can scale up the size of the subject and at the same time reduce the radiofrequency power requirements and increase the sensitivity.

The concept of phased array coils in MR imaging was introduced by Roemer et al. (1) as an efficient approach to improve signal-to-noise ratio (SNR). In this approach, multiple smaller resonators, decoupled from each other, each covering a specific field of view (FOV) is used in place of a larger single resonator. Because smaller resonators will have superior sensitivity, though covering a smaller fraction of FOV, combining data from multiple resonators can in fact lead to overall improvement of SNR. The individual resonators can be operated simultaneously with as many receiver chains, or can be multiplexed and measured in an inter-leaved fashion. Such an array concept has now become very popular and modern MRI machines which have up to 64 or 128 receive channels to enable the use of arrays with large number of individual coil elements. This has also led to the development of parallel imaging modalities with special pulse sequences and partial k-space strategies to speed up the throughput of clinical imagers (2–6). For small animal imaging, the use of phased array coils can replace the conventional bird-cage or other similar resonators (7,8).

While the use of array resonators has now become routine in several MR imaging applications, such strategies have not been attempted in time-domain electron paramagnetic resonance (EPR) imaging. Time domain EPR imaging, although conceptually similar in principles and methodology to MRI, was challenging to implement for in vivo applications because of the microsecond range relaxation times of the paramagnetic systems in comparison with that of the nuclei whose relaxation times are in the millisecond to seconds range. Therefore, nanoseconds excitation pulses, switches with nanoseconds time resolution for rapid isolation, and quenching of transmit power are needed to minimize the resonator ringing which can otherwise outlive the duration of the free induction decay which typically last for a few microseconds. Resonators with very low Q and T/R isolators (diplexers) with fast switching in ~100 ns range are used in most systems. Apart from this, attempts to spatially encode paramagnetic systems with even mm resolution in small objects, few cm in diameter, require uniformly exciting and capturing frequency bandwidths in the order of tens of MHz which is a challenge. Pulse widths in the order of tens of nanoseconds and power levels of several hundred Watts are necessary. Small animal FT-EPR imaging modalities have been developed in our laboratory and elsewhere (9–16) in the time-domain mode and routine use of EPR imaging is being made in mouse models of tumor to study and map tumor hypoxia (regions of tissue where the pO₂ is less than 5 mmHg). The importance of mapping out tumor hypoxia stems from the fact that hypoxic tumor cells are about 3–4 times more resistant (17–19) to radiation and chemotherapeutic killing compared with normal cells, and a mapping of tumor hypoxia can help plan radiotherapy based on pO₂ levels and improve the prognosis of radiation treatment of cancer. In order to apply EPR imaging to larger animals and ultimately to humans, it is necessary to scale up the resonator volume/size. It will also be necessary to use much higher power (~kW) and much smaller pulse widths (~10 ns) which may lead to unacceptable specific absorption rate (20). It is in this context that we believe that array resonators will help make the transition from small animals to human subjects and topical imaging of tumors by Fourier transform electron paramagnetic resonance (FT-EPR). Array coils have been used effectively in MR imaging of the entire spinal cord, for example, in the study of multiple sclerosis (21). In this study, we describe the use of planar surface coil array of 22 mm diameter as well as a pseudo volume resonator formed by four surface coils with adjacent pair arranged orthogonally. The results obtained demonstrate the feasibility of using such resonator configuration in FT-EPR imaging and increase the size of the object being imaged without significantly increasing the rate of radiofrequency (RF) power deposition. Because each array element needs to transmit and receive only fractional FOVs, the individual bandwidth coverage is also reduced enabling the use of lower power and reasonable pulse widths.

METHODS

300 MHz Spectrometer/Imager

The 300 MHz FT-EPR imager operating in our laboratory and the way in which we perform EPR imaging have been described in detail previously (22). In brief, it consists of 10 mT electromagnet with an effectively uniform volume of 5 cm³ field with field homogeneity of 10 ppm. Cylindrical parallel loop resonator designed and patented by us (22–26) with sizes 25 mm diameter × 25 mm length and 25 mm diameter × 50 mm length are routinely used for mouse imaging. The magnet is also equipped with a three-axis gradient system capable of a maximum gradient of 40 mT/m in any direction. Single point imaging modality is used for image formation (9) where the system is subjected to pulses of 70 ns duration at 200 W and the free induction decays are collected in presence of gradients that are stepwise ramped from a −G to +G in ΔG steps. The magnetization measured at a particular time delay from the pulse in monitored as a function of rastering the gradients in either a plane (for 2D) or along all the Cartesian axes (3D) forms the k-space which upon 2- or 3-dimensional FT yields the image. Because the entire free induction decays are collected one can generate as many images as one needs as a function of delay from the pulse until the SNR becomes too low. The FOV of images are given by

$$\text{FOV}=2\mathrm{\pi }/(\mathrm{\gamma }\mathrm{\Delta }G\mathrm{\tau })$$ [1]

where γ is the electron gyromagnetic ratio, ΔG is the gradient step, and τ is the delay from the trailing end of the pulse. An increase in the number of steps will increase the image resolution proportionately. It will be possible to use a sequence of images as a function τ and generate images of identical FOV (by interpolation or other numerical methods (27)) with intensities weighted by the apparent relaxation time $T_2^{*}$ from which one can obtain the pO₂ because $T_2^{*}$ linearly depends on pO₂. The way in which quantitative oximetry is performed using FT-EPR and single point imaging has already been detailed previously (22,25).

The 300 MHz EPR imager used in this study is described earlier (11,22) and was used for the phantom and in vivo experiments with array coils. The array coil resonators (surface loops) used is illustrated schematically in Figure 1. The resonator typically consisted of three or four surface coil elements. When a part of the resonator circuit is manually shorted by conductive wire, the resonance frequency of the resonator was shifted away from 300 MHz thereby suppressing mutual coupling between adjacent coils.

FIG. 1.

Scheme of sequential pulsed EPR image acquisition. Surface coils were selected by sequentially connecting them one by one to RF source, whereas the others were shorted which let to frequency offset and minimization of coupling.

Individual Surface Coils

Figure 2 illustrates the configuration of the surface coil resonator. The resonator consists of a surface coil, parallel coaxial lines, and a balun constructed by half-wave line (28–30). Frequency adjustment was achieved by inserting the capacitors in series to the transmission line and changing the capacitive reactance using trimmer capacitors which are illustrated as C_T in Figure 2. Impedance matching adjustment was also achieved by same way as the frequency adjustment using trimmer capacitors which are shown as C_M1 and C_M2 in Figure 2. While Q was lowered mainly by over coupling, further reduction was achieved using a resistor in parallel to the transmission line. The surface coil was made from 18 AWG tinned copper wire. The coil diameter was 22 mm and the distance between centers of the coils was 16.5 mm to minimize coupling between each other. To electrically insulate the coils from each other, the coils were covered with heat shrink tube. The coupling circuit was made from copper laminate substrate.

FIG. 2.

Configuration of individual surface coil at 300 MHz. Parallel transmission line and a half-wave line balun were formed with 50-Ω coaxial lines. Trimmer capacitors were adjusted by adjusting the impedance matching between the resonator and the transmission line connected to the bridge. 560-Ω resistor was used to reduce the quality factor Q and the ringing time of the resonator.

Nonmagnetic chip capacitors (American Technical Ceramics, Huntington Station, NY) and trimmer capacitors (Johanson Manufacturing, Boonton, NJ) were used in both tuning and impedance matching. To shift the resonant frequency of the “passive” coils, the end of the parallel transmission on the side of a coupling circuit was shorted as mentioned previously causing a considerable shift in the resonance frequency. The parallel transmission line and the half-wave balun were made from coaxial semi-rigid cables (model UT-141C-form, UT-141C-TP, Micro Coax, PA).

Chemicals

Triarylmethyl (Oxo63), methyl tris(8-carboxy-2,2,6,6-tetrakis-(2-hydroxyethyl)-benzo[1,2-d:4,5-d′]bis(1,3)dithiol-4-yl), procured from GE Healthcare (Wauwatosa, WI) was used for all experiments in this article.

Animal Preparation

Female C3H mice were obtained from the Frederick Cancer Research Center, Animal Production (Frederick, MD 21702). The mouse was 12–16 week and weighed 27.2 g. Squamous cell carcinoma (SCC-VII) cells were implanted in the right hind leg of mouse and grown in 10 days. The tumor size was ~12.4 mm width, 14.2 mm length, and 11.0 mm thickness. The mouse was anesthetized using isoflurane® (induced at 3.0% and maintained at 1.5% in 750 mL/min medical air). Oxo63 (75 mM) was given as a 1.125 mmol/kg bolus followed by 0.04 mmol/kg/min continuous injection via tail vein cannulation and imaging was started immediately thereafter. Experiments were carried out in compliance with the Guide for the Care and Use of Laboratory Animal Resources (1996), National Research Council, and approved by the National Cancer Institute Animal Care and Use.

Phantom Imaging

Figure 3a,b shows the photographs of the resonator made of three coils elements and the phantom for 2D imaging. To investigate the area of visualization of the surface coil array, we constructed a plane surface coil array using three resonators and measured 2D EPR images of a flat rectangular cuvette-like cell phantom (22 × 52 × 10 mm³) filled with 3 mM Oxo63 aqueous solution. The parameters for 2D EPR imaging were as follows: FOV = 75 × 75 mm; maximum magnetic field gradient = 6.0 mT/m; applied microwave power = 100 W; number of projections = 361 (19 × 19 steps along X and Z directions in 0.66 mT/m steps); number of transients averaged = 10,000.

FIG. 3.

Photographs of surface coil array and the phantoms. a: Three-channel coil array for 2D imaging, (b) the rectangular cuvette-like phantom filled with Oxo63 solution, (c) four-channel coil array configured around a cylindrical polyethylene tune to simulate a cylindrical volume coil, (d) a syringe phantom filled with 3 mM Oxo63 solution, and (e) overall view of the surface coil array. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Figure 3c,d shows the photographs of the surface coils and the phantom used for 3D imaging, and Figure 3e shows the resonator with four individual coils with their tuning and matching circuits. To obtain 3D images, we configured the surface coil array like a volume coil using four resonators on the surface of a cylindrical polyethylene tune fixed with adhesive tape, and measured 3D EPR images of a cylindrical phantom (14.8 mm in diameter and 24 mm in length) filled with 3 mM Oxo63 aqueous solution. In 3D imaging with the surface coil array, four individual surface coils were placed and fixed with adhesive tape around a cylindrical plastic tube (21 mm in diameter). The parameters for 3D EPR imaging were the same as mentioned above for 2D imaging, except with 6859 projections (19 × 19 × 19 steps along the X, Y, and Z directions in 0.66 mT/m steps) and an FOV of 75 × 75 × 75 mm³. In vivo measurements of the mouse leg tumor were also carried out by placing the tumor-bearing leg in the array coil assembly and using the same parameters. The 2D and 3D images were obtained by FT of k-space generated from a single time point at a delay of 700 ns from the pulse, after zero-filling to a matrix size of 64 × 64 and 64 × 64 × 64, respectively.

RESULTS

Phantom Imaging

To test the feasibility of using a set of array coils to image an object, which is larger than a component single coil, we have performed both 2D and 3D imaging on phantoms. For 2D imaging of a large sized phantom depicted in Figure 3b consisting of a flat rectangular glass cell filled with 3 mM Oxo63 solution was used. The diameter of the individual array coil was 22 mm, and three coils were arranged in a plane overlapping with each other over a region of 5 mm (Fig. 3a). The three coils were insulated from each other and each coil was used to capture the image from the phantom within its RF flux domain. Gradients were applied in a Cartesian raster parallel to the plane of the array in steps of 0.6 mT/m in the range +6 to −6 mT/m. Figure 4a shows the 2D EPR images of the flat rectangular cell obtained by single point imaging using the individual resonators in XZ-plane. Figure 4b shows the combined EPR image measured with the surface coil array. Before combining 2D EPR images obtained by individual resonators, they were normalized due to the differences in sensitivity between coils. The area of visualization in the individual images of Figure 4a was approximately equal to the inner area of each coil, whereas the combined image Figure 4b had the FOV larger than the phantom. The FOV of the image was 72 mm × 72 mm. Thus, it is possible to expand the effective FOV of the image to the total array area enabling larger objects to be imaged. Further, it is obvious that the power required should be much less than what would be required if single large coil was used.

FIG. 4.

2D EPR images of a rectangular phantom with a “three-channel” surface coil. a: Images from individual coils and (b) the combined image with scale factors based on the relative sensitivity of the coils.

To test the feasibility of using a combination of array coils to perform 3D imaging instead of a volume coil, we used four surface coils placed on the surface of a cylindrical tube (Fig. 3c). Figure 5a–d shows the surface-rendered 3D images of the phantom tube obtained by individual resonators. Figure 5e–g shows the surface-rendered image combined from 3D EPR images measured with individual coils for various combinations. Although the individual images were not scaled, the combined image (Fig. 5g) generated a smooth cylinder of dimension in satisfactory agreement to that of the actual phantom. Figure 5h–j shows the slices from the 3D image along XY, YZ, and XZ planes generated from the combined image with dimensions correctly reproducing the distribution of the spin in the phantom. It is apparent from the figure that the RF flux does not reach far enough from the coil plane leading to a small cylindrical hole (indicated by an arrow) on the center of the combined image. This can be corrected either by scaling the individual images radially or by using a coil of larger radius. Nevertheless, this data suggests that it is possible to scale up EPR images to larger subjects by using multiple array coils.

FIG. 5.

EPR images measured from the tube phantom with four-channel surface coil array. a–d: Surface-rendered images obtained with individual coils. Combined surface-rendered images [(e) is the sum of (c and d), (f) is the sum of (c, d, and b), and (g) is the sum all four individual images. Slices generated from the 3D image along mutually orthogonal axis are shown in (h–j). The central hollow region (see text) is indicated by the arrows.

In Vivo Imaging

Using the coil arrangement described above, 3D imaging of the tumor-bearing leg of C3H mouse was carried out. The positioning of the mouse leg in the Lucite enclosure surrounded by four surface coil resonators is shown in Figure 6a. 3D image data was collected sequentially using each individual coil with the gradient increment step of 0.66 mT/m steps from +6 to −6 mT/m in 19 × 19 × 19 Cartesian raster. The free induction decay data from each coil were processed according to the procedure used for single point imaging. The surface-rendered 3D images from the individual coils are shown in Figure 6b and a combined image obtained by summing the individual images is shown in Figure 6c. Because the individual images are all measured under identical gradient settings taking care that the FOV covered by each element is large enough to cover the entire object, the resolution of the individual and the combined images are expected to the same.

FIG. 6.

EPR images of an SCC tumor bearing leg of a C3H mouse, measured with a surface coil array of four coils positioned on the surface of cylindrical tube mimicking a cylindrical volume resonator. a: Schematic showing the positioning of the tumor-bearing leg inside the polyethylene tube surrounded by four surface coils. b: Surface rendered 3D images from the individual coils. c: Combined surface-rendered 3D image. d: 2D slices of images generated from the combined image.

Figure 6d shows a few slices through the composite 3D image of the mouse-leg tumor. These slices resemble what would be obtained from an actual image using a single volume coil suggesting that it is possible to perform 3D imaging using suitably configured surface coil array. Eventually, one can multiplex the coils or inter-leave the measurement using power levels that would be much less than what would be required if a single volume coil is used. The main motivation for this work is to demonstrate that if one has to scale up EPR imaging to larger subjects, the way to reduce specific absorption rate would be through the use of surface coil array and combining the images after suitable scaling. Further work is in progress to model the actual RF flux profile and to generate 3D scale factors that can reproduce the spin density correctly and enable T₂ or $T_2^{*}$ based quantitative oximetry.

DISCUSSION

It is possible to use a combination of smaller surface coils to image a larger object compared with a single coil. It is also possible using appropriate configuration of the coils to emulate volume coils in an attempt to image large objects. This capability is useful in time-domain EPR imaging because the short transverse relaxation times of paramagnetic spin probes used for imaging (trityls, deuterated nitroxides, LiPc and derivatives) compels the use of resonators of low Q that will help shortening the resonator ring down time and allow capturing the time response within a fraction of a microsecond. On the same ground, it is also important to use pulses of 20–80 ns duration to cover the band width to be excited and detected even under moderate spatial encoding gradients. With the line widths of the spin probes on the order of 0.5 MHz, one needs to cover tens of MHz to have meaningful image resolution. This further imposes additional requirements on the power to be packed in ultra-short pulses requiring 200–500 W amplifiers, depending on the volume of the resonator, even for small animal imaging. Scaling up of the imaging object (and hence the resonator volume) brings with it further increase in required pulse power leading to unacceptable specific absorption rate. Power reduction strategies using stochastic excitation (31,32) or specialized chirp sequences (33) have been tried in MR imaging. Another practical way to alleviate the increase in power requirements and at the same time scale up the object size (FOV) is by the use of multiple resonators, also a well-known development in the MRI field (1). These coils may be phased array coils which are multiplexed and used sequentially in a time-shared fashion or in parallel using multiple receiver channels. The use of multiple array resonators is likely to be a solution to accomplish pulse EPR imaging of large objects and eventually human anatomy with acceptable specific absorption rate levels. The feasibility that is demonstrated here may well encourage extension of such a strategy to make time-domain EPR imaging a viable in vivo molecular imaging modality.