Multimodal imaging is of great interest due to the possibility of combining different sources of contrast¹. In particular, US and MR imaging are often manually combined in normal clinical practice to take advantage of their complementary contrast mechanisms. A combination of US and MR imaging may help to increase the true detection rate of cancer², cardiovascular disease (which accounts for 32% of all U.S. deaths³), and other conditions. However, we are unaware of any low-cost point-of-care or wearable system that integrates both imaging modalities. We propose a fundamentally new design; a miniaturized 2-D US sensor collocated with a 1-D, low-field MR sensor for bimodal imaging in portable or wearable form factors. The end goal is to develop a system in which the US and MR sensors, electronics, and signal processing unit are contained within one instrument. This paper describes work done to experimentally verify the feasibility of such an integrated US and MR imaging system.

MR sensor: The single-sided low-field MR sensor used in this work has been previously described⁴. It uses an array of 3 low-cost permanent magnets and has a usable depth range of 4-12mm. External RF interference is reduced by using a butterfly-shaped planar RF coil (gradiometer). The typical Larmor frequency and vertical resolution at a depth of 7 mm are 8.26MHz and Δy≈94µm, respectively. Figure 1 shows a block diagram of the sensor. A benchtop NMR spectrometer (Kea2, Magritek) is used to generate the pulse sequence and collect data. Two-dimensional relaxation-relaxation and diffusion-relaxation maps (T₁-T₂ and D-T₂) were measured at various depths within the skin of an adult volunteer using the setup shown in Figure 2(a). The sensor was also used to measure the velocity of a sample (doped water) flowing through Teflon tubing.

US-MR integration: We used the single-sided MR sensor and a commercial US sensor to collect data from the same phantom. Four PEEK cylinders (OD=1.7mm) are arranged inside a glass tube (ID=4.1mm) filled with PBS. This phantom is placed inside a silicone rubber mold to simulate human tissue. The phantom arrangement is shown in Figures 3(a) and 3(b). B-mode 2-D US images (xz-plane) were measured by a commercial medical-grade US imaging system (Risingmed RUS-9000B) using a linear transducer array (6-8.5MHz). 1D MR images (y-axis) of the same sample were obtained by using a computer-controlled motion stage to move the sample along the y-axis. The experimental setup described above can be seen in Figure 2(a).

MR experiments: Figure 2(b) shows measured results from human skin. The D-T₂ maps show that bound water decreases with skin depth while unbound or free water increases. Figure 2(c) shows the results of the velocity experiment. The extra relaxation caused by sample motion increases linearly with flow rate, so the mean velocity of samples with known T₂ can be estimated without a pulsed field gradient system.

US-MR experiment: Figure 4 shows results from the combined US-MR experiment. The bottom left shows the US image, with the tube being the bright line running from top to bottom. The graphs on the right show the MR results (initial signal amplitude and T₂ of the PBS). The US image clearly shows the exterior of the glass tube but not its internal structure; the MR scan measures properties of the liquid (PBS) and resolves fine structural details within the tube, including periodic reductions in signal amplitude (but not T₂) caused by the PEEK cylinders. This result shows the effectiveness of low-field MR in imaging objects that are inside acoustic shadow regions and thus invisible to US.

Our experimental results demonstrate the feasibility of portable or wearable integrated US-MR imaging. The combined data shows that US has larger penetration depth, while MR has greater depth resolution and can image within acoustic shadow regions. We are currently developing a custom linear US transducer array that is physically thin enough (thickness <4 mm) to be mounted underneath the magnet. We are also developing a set of planar coils to apply pulsed field gradients along the x and z directions, which will allow three-dimensional (3-D) imaging with the single-sided MR sensor⁵. The end goal is to create integrated US-MR probes for autonomous diagnosis of tumors and cardiovascular disease by developing visualization tools to fuse 2-D US and 3-D MR data, and also machine learning algorithms to classify the fused images. Figure 5 shows a conceptual design for such an integrated probe. The results of this paper show that such a system has the potential to be a valuable tool for studying the properties of tissue.