3D-printed gear system for antenna motion in an MR environment: initial phantom imaging experiments

Abstract

We have developed a microwave imaging device for breast cancer imaging that can be used concurrently inside an MR imaging system. The microwave measurement system is comprised of a horizontal array of 16 monopole antennas that can be moved vertically for full 3D coverage of the breast. All compatibility issues have been addressed. The motion is achieved using a novel 3D printed gearing device. Initial results demonstrate that the system is capable of accurately recovering the size, shape, location and properties of a 3D shape varying object. This is a critical step towards clinical microwave breast imaging in the MR.

I. Introduction

MR is a powerful modality for many medical applications, but comes with important caveats depending on the situation. Cost and accessibility are clearly issues at a global level. However, for certain indications, it is simply the best option. For diagnosis of breast cancer, especially for women with dense breasts, it is one of the leading options [1]. While MR provides excellent spatial resolution, its ability to differentiate soft tissues is limited [2]. This is particularly the case for breast cancer diagnosis in denser breasts. To overcome this limitation, radiologists routinely use contrast agents and monitor the drug uptake and washout behavior to quantify the response. While this is effective, there are numerous benign tissues which enhance the images in the presence of gadolinium and lead to a large number of false positives. This ultimately leads to increases in patient inconvenience and health care costs. In addition, more recently, there has been increased concern about the health risks in using gadolinium. While still considered safe, there are now documented examples of trace amounts of the contrast agent localizing in the brain [3]. Previous safety concerns were allayed because it was assumed that the gadolinium was cleared from the system relatively quickly. These issues have raised interest in exploring alternatives to contrast agents – particularly non-invasive techniques.

Microwave imaging is poised to fill this role. It presents good endogenous contrast in the dielectric properties between normal and malignant breast tissue [4]. However, as a stand-alone modality, it is not as refined as MR and others regarding spatial resolution due to the smoothing aspects of the associated image regularization processes. More recently, we have developed a numerical technique that can fuse the advantages of the MR spatial resolution with the specific nature of microwave imaging – called the soft prior regularization [5]. Obviously, these results are best when both the MR and microwave exams can be performed concurrently to ensure perfect registration of the images. In earlier work we were able to successfully demonstrate a simple prototype that allowed for microwave imaging a single plane of a patient’s breast while the patient was in the MR [6]. This provided the excellent spatial resolution via the MR data while substantially improving the discrimination of the tissue types via the specific nature of the tissue dielectric properties. It is important to note that the microwave imaging data was not biased by any assumptions regarding the dielectric properties of the different tissue types. The algorithm only started at those properties of the surrounding bath.

While promising, the greater challenge remains as to microwave imaging the whole breast volume while inside the MR bore. Per our previous hardware projects, it is still useful to perform the microwave imaging utilizing a hybrid of a set of monopole antennas and then exploit mechanical motion of the array to increase the amount of measurement data for full 3D reconstructions. The two most prominent challenges are to provide the motion in such a constrained space and not use magnetic or incompatible metallic parts in the setup. Other challenges include minimizing multipath signal corruption and eliminating the risk of liquid spills, both of which we have considerable experience. The monopole antennas are ideally suited for this situation. Because of their low profile and minimal amount of metal, we will be able to incorporate baluns on the feedlines so that they will have negligible effect on the MR images. Our first prototype is a mechanical gearing system utilizing a combination of a worm gear concept with a fairly standard acme screw to convert horizontal rotation motion into vertical motion of the antenna mounting plate. Both of these gearing systems can be fabricated to not dramatically add to the overall height. In this case we were able to design and fabricate the parts using 3D printing technology that provided uniform array motion and sufficient strength to push the coax feedlines through the tank base hydraulic seals. Figure 1a shows a SolidWorks (Waltham, MA) rendering of the tank and mechanical motion system (antenna feedlines not shown) and lb is a photograph looking down into the imaging tank showing the monopole antennas and some of the gearing mechanism through the acrylic tank base.

Figure 1.

(a) SolidWorks side view of the imaging tank and gearing mechanism, and (b) photograph looking into the tank showing the array of monopole antennas and the gearing system underneath the tank.

II. Results

There are two important aspects to the data that need to be mentioned – the MR images and the microwave images. First, the object being imaged was a solid cone made out of a fat-equivalent material. It is 8.1 cm high with diameters of 1.45 and 4.55 cm at the top and bottom, respectively. The MR images are able to accurately recover the objects, albeit, there are artifacts in the vicinity of the semi-rigid coaxial feedlines (Fig 2). The disruptions are small and have no impact on the resolution within the microwave imaging zone. These will be eliminated using baluns which will be implemented at a later date. The microwave images are taken at four levels from near the top to the bottom. The imaging algorithm utilizes a Gauss-Newton iterative scheme with a log transformation for the variance stabilization [7]. Once the transformation has been implemented, we utilize a standard Tikhonov regularization. The process does require phase unwrapping which is performed via a frequency reference for the measurement data and with a normalized iteration reference for the computed phases [8]. Figure 3 shows the recovered object at one plane. The 2D algorithm typically converges in less than 20 iterations and with the new discrete dipole approximation-based scheme, takes under 6 seconds [9].The object is in the correct location and has the correct properties and diameter.

Figure 2.

MR images of the microwave illumination zone. Right image shows the artifacts along the horizontal portions of the microwave feedlines.

Figure 3.

1300 MHz reconstructed images – permittivity (left), conductivity (right) of the fat equivalent phantom.

III. Discussion and Conclusions

These results demonstrate a first recovery of a 3D varying object while inside an MR bore. Most of the more minor issues have been addressed including reduction of multi-path signals and elimination of bath leakage. Both MR and microwave images are sound. We have not yet fully implemented the 3D imaging algorithm for these phantom experiments, but the recovered 2D images clearly demonstrate the correct progression in shape change as a function of vertical position. The next generation system will incorporate specialized RF coils to improve the signal-to-noise ratio and the magnetic field uniformity for improved image quality. Additionally, the next generation system will be substantially shorter to allow an actual patient to be imaged. These results are encouraging and set the stage for actual clinical testing.