Menu

Wireless Sensors For MRI

Motivation

This project is is a Skunk Works effort I began as a postdoc at Stanford. Several people in our group work on correcting MRI images for motion during an MRI scan. Generally, information on motion is obtained from the image itself, and we thought it may be interesting to create an alternate means to acquire this information. I devised a system that employed a wireless array of accelerometers to measure the shape of the chest (see the conference abstract below for discussion of the measurement concept) and send it back to the control room in real time. As I write this, I'm working on a version with a PLL on board which will harvest a clock at a different frequency, and determine the relative phase of the wire current.

Soon after building and testing the initial array, other problems began to surface that might be addressed with a wireless sensor. I built a model to measure the current in a wire, which is important thing to quantify in our safety program, and also one to measure the heartbeat. Effective cardiac triggering is crucial in some applications, and alternate methods are required in ultra-high field systems.

Generations

XBee, Arduino
I got on to all of this by reading an article in Elektor on some XBee-based project. At this point I had very little knowledge about things like this, so I got started easy with an Arduino and an XBee radio. The result is pictured on the right, where there are three accelerometers on a band that straps around the patient (this project involves sewing!). There's a shielded cable that runs back to a base, which contains the Arduino and the XBee. This setup worked fairly well, but seemed to generate some interference in the image. This led me to the next generation, which is completely wireless. This stage, however, taught me a lot about the shielding requirements of such a hostile environment.

The first prototype of my sensor array, using XBee radios and Arduino
ATMega128RFA1
The images were a bit noisy in the first experiments, and to be honest, the whole thing was a bit bulky. The obvious thing to do was to make the system truly wireless! I built a new system around the ATMega128RFA1, which has an on-board radio transciever, running at 2.4 GHz. Pictured on the right, it also has a custom mount which I made with a laser cutter. The non-magnetic battery sits between the circuit board and the mount.
My second generation sensor, outfitted with a current-sensor
Test Setup, Base Station
Here is a picture of my test setup. On the left is a sensor board with the radio and MCU, but without the peripherals. Instead of a chip antenna, it has a large duck antenna which is attached to an SMA connector. There is an SPI link over to my Beaglebone Black, which then fires the data off across the network. You can also see a couple of the sensors on the bottom, and also a development board for the ATMega128RFA1 from Sparkfun. All the spaghetti coming out of the Beaglebone is for another project.
Test setup for the wireless sensors
Heartbeat Sensor
This version uses a satellite PCB to sense the heartbeat
This version uses a satellite PCB to sense the heartbeat
Current Sensor with Phase
I've designed this, but haven't built it yet!
I've designed this, but haven't built it yet!


Current Sensing Prototype - Conference Abstract

A Completely Wireless Current Sensor for RF Safety
C.W. Ellenor(1), P.P. Stang(1,2), M. Etezadi-Amoli(1), J.M. Pauly(1), G.C. Scott(1)
(1) Department of Electrical Engineering, Stanford University, Stanford CA, USA
(2) Procyon Engineering, San Jose, CA, USA


Introduction One of the main remaining hurdles in MRI safety is the interaction of imaging fields with long wires, which may be part of a medical device such as a pacemaker or neurostimulator, or may take the form of a guidewire or catheter in an interventional procedure. The RF imaging field is capable of depositing large amounts of energy in such wires, and tissue damage may result, precluding a large segment of patients from receiving MRI scans. The goal of quantitative, real-time monitoring of these interactions holds the promise of both increased safety during a scan, and more systematic evaluation of medical device interactions. We believe that physical measurements are key to systematic assessment of both risk, and of safety measures. To measure induced RF current, a sensor which fits over an exposed segment of wire has been demonstrated in [1]. This sensor is toroidal in shape and couples flux from the region around the wire to produce a signal. The received signal is used to modulate a photodiode, which transmits the signal via optical fiber. In further work [2], the battery has been replaced by a photonic power supply and, and has been validated by an imaging study. In this abstract, we demonstrate a completely wireless version of this sensor, which is powered by a non-magnetic LiPo battery (Powerstream Technology, USA), and where the signal is digitized on board and transmitted using an 802.15.4 2.4GHz radio. A wireless link avoids the added infrastructure, inconvenience and risk of additional optical or coaxial cabling.


Methods The signal from the toroidal detector is received by an RF power detecting IC (LTC5507, Linear Technology, Milpitas, CA), which detects power from -34dBm to 14dBm in the range of 100kHz to 1000MHz. The output returned by the LTC5507 is fed to a 10-bit ADC, which is on board an Atmel ATMega128RFA1 microcontroller, which has an integrated radio transceiver capable of up to 2 MB/s data transmission. Samples are acquired at 10ks/s, though rates up to 330ks/s are supported. The measurement is triggered wirelessly by a second transceiver, which is located in the control room and also receives the acquired data. It is connected to an antenna in the scanner by a coaxial cable. For a demonstration, we have created a wire phantom using a length of wire inside a tube containing saline solution. The wire protrudes from the end of the phantom, and the sensor is placed over this wire. The wire phantom is placed inside a birdcage coil whose two drive points can be driven separately, effectively creating a phased array. The array configuration which minimizes current on the wire is determined by sweeping the amplitude of one channel up while the other is swept down and the phase between channels is varied rapidly. This effectively samples the parameter space of the array, and is illustrated in the lower panel of Figure 3.


Results and Discussion The upper panel of Figure 3 shows the result of the current measurement described above. The radio transceiver timing is based on a 16 MHz crystal oscillator, giving rise to concern about interference in imaging. It was found that when the radio was placed within a few centimeters of the receive coil, some extra signal was observed at a single frequency, but in most images was undetectable. We expect that further improvements in shielding can alleviate this concern. Because an RMS detection scheme is used, phase information of the measured current is not recoverable from the sensor. Despite this, using the scan shown, the sensor can be used to deduce the relative amplitudes and phases of contributions to wire current of any two elements of the array, and can therefore be used iteratively to construct null modes even in more complex systems.

Conclusion A new, completely wireless current sensor has been demonstrated for the detection and monitoring of RF current in wires. Its small form factor and lack of cables make it a valuable tool in the typically congested MRI environment.


References [1] Zanchi et al., IEEE Trans. Med. Imag. 29:169-178, 2010. [2] Etezadi-Amoli et al., Proc. 21st ISMRM, p723, 2013 NIH Grant Support: R01EB008108, R33CA118276, R21EB007715, P01CA159992.



Motion Sensing Prototype - Conference Abstract

A Wireless Accelerometer Array for Respiratory Motion Tracking
C.W. Ellenor, N.O. Addy, R.R. Ingle, D.G. Nishimura, J.M. Pauly, G.C. Scott, P.P. Stang
Department of Electrical Engineering, Stanford University, Stanford CA, USA
Purpose. Motion artifacts are a well-known problem in MR imaging, which appear as ghosting or blurring in reconstructed images based on the imaging trajectory. A common approach to reducing this artifact is the inclusion of independently measured displacement information, typically by navigators or probes, into the reconstruction algorithm. In this abstract we present an accelerometer-based system for measuring displacements of the thorax during respiration. Our system differs from previous accelerometer-based attempts [1] in that we use an array of accelerometers measuring tilt to reconstruct the instantaneous shape of the chest, rather than doubly integrating the acceleration signal at a single point. We find this approach to be more sensitive and more robust against noise. Because our sensor can measure absolute displacements and rotations in real time, we expect it to be valuable for a wide range of prospective and retrospective gating and motion correction techniques. In addition, we have developed a wireless system to transmit data from the scanner to the control room, eliminating the hazard and inconvenience of long cable runs.
Figure 1 - The accelerometer array with band. The box contains the MCU and 2.4 GHz radio.
Methods. Figure 1 shows a picture of the prototype system. An elastic band with Velcro wraps around the patient, holding the 3 accelerometers in place. A shielded wire connects the accelerometers to a box containing a microcontroller and a 2.4 GHz XBee radio transmitter, which relays data back to the control room via an antenna mounted just outside the bore. A non-magnetic lithium polymer battery powers the unit. We model the thorax as an ellipse [2] where it is the length of the semi-minor axis a that we will attempt to measure (Figure 2A). For a fixed sensor at angle θ along the circumference, the tangential angle φ and the parameter a are related as
The tangential angle is measured by an accelerometer as the ratios of the projections acceleration due to gravity onto the three axes. Figure 2B shows the magnitude of the signal expected along two accelerometer axes for a tilt-based measurement and an integration-style measurement. When combined with an elliptical model, the measurement of tilt provides a measurement several times more sensitive than would a doubly integrating single accelerometer. The combination of multiple axes and devices will also suppress common mode noise sources such as drift. A healthy volunteer was fitted with the accelerometer array about the abdomen. Nominal inclinations of the accelerometers were ± 60 degrees, and scans were run on a 1.5 T Signa Excite scanner. A 2-D spiral, gradient-echo, cardiac triggered scan was performed with sagittal and axial navigator images taken before and after each heartbeat, respectively. The 28 cm FOV, 3.1 mm resolution, 12- interleaf 2D spirals were acquired with TE/TR = 1.6/6.3 ms, acquiring a single image in 75.6 ms. Approximately 83 images per minute were acquired for about four minutes. The location of the chest wall was determined from each image by fitting a peak to a line profile taken through the center of the axial image, along the A/P direction. The position was compared to that measured by the accelerometer array, which was sampled at approximately 30 Hz.
Figure 2 - Measurement concept (A) geometry (B) sensitivity compared to double-integration for 1cm displacement at 0.5 Hz
Results. Figure 3B shows the results of the scan. The chest displacement determined from the accelerometer outputs displays excellent agreement with that determined from the navigator images (correlation coefficient = 0.88), though some discrepancy in amplitude does exist. The images studied (Figure 3A, for example) show artifacts which are generated by the device. Our hypothesis is that these artifacts are caused by field inhomogeneities due to the wire shielding, and we expect to be able to eliminate these in subsequent generations of the device. Our architecture supports further miniaturization and even the complete elimination of wires.
Discussion. Though we have aimed this demonstration at the recovery of a single point on the chest wall, it is clear that more detailed geometries can be recovered by increasing the number of sensors and the sophistication of the reconstruction model. We have built the system to be inherently scalable, and envision future versions to employ a band or bands of many sensors.
Conclusion. We have prototyped a new, wireless system for recording quantitative data on the changing shape of the chest during respiration, and demonstrated its effectiveness using navigators. Future work will expand the size of the array, and develop models for determining motion of different anatomy. We will also investigate the utility of these signals for respiratory and cardiac gating.
References. [1] Rousslet L, et. al. ISMRM 2011; 4601 [2] A. De Groote et. al. J. Appl. Physiol. 83; 1531
Figure 3 - Results (A) navigator image used to locate chest wall. Arrow indicates location of profile. (B) Position of chest wall as determined by navigators and accelerometer array.

Some Things I Learned While Doing This

- PCB layout
- Microcontroller programming
- SPI, I2C protocols - QFN Soldering
- RF shielding of circuits
- EE 802.15.4
- Laser Cutting