Wii System Components to Track Ultrasound Probe and Create 3-D Images

Investigators: Van Gambrell, Laura Owen,

Steven Walston, Jonathan Whitfield

Advisor: Christopher Lee, M.D.

Date Completed: April 27, 2010

ABSTRACT

Amyotrophic Lateral Sclerosis (ALS) is a neurodegenerative disease that progressively destroys motor neurons and causes atrophy in skeletal muscle. The goal of our project is to integrate an ultrasound imaging method with tracking techniques from Nintendo’s Wii system to obtain accurate skeletal muscle volume measurements. Three tracking systems were created each returning x, y, and z location along with roll, pitch, and yaw angle orientation. One system involved the IR camera tracking an LED array, the second system added the WiiMotionPlus to return yaw, pitch and roll, while the third system used the accelerometer from a WiiMote to give x, y, and z with the WiiMotionPlus. The data from each system was then implemented into a program that aligned and rotated every US slice to give a final 3-D reconstructed image. The error from all three systems was greater than the error of the US machine, producing unusable 3-D images. A more accurate tracking system could be implemented to improve results.

INTRODUCTION

Amyotrophic Lateral Sclerosis (also referred to as ALS or Lou Gehrig’s Disease) is a neurodegenerative disease that currently affects 20 - 30 thousand people in the United States. The disease targets and progressively destroys spinal motor neurons that stimulate skeletal muscles [1, 2]. Muscle fibers require neural stimulation to maintain strength and structure. As a result of ALS, the damaged spinal neurons and their respective nerve fibers are unable to send signal impulses to muscle, effectively causing the atrophy of muscle [3, 4]. This phenomenon is diagramed in Figure 1. As the disease progresses, more neurons become diseased and waste additional skeletal muscle fibers. Over time, what was once full, normal muscle atrophies into a thin, weak strain of fibers.

The full body degradation of muscles eventually causes semi or full paralysis in patients accompanied by speech, breathing, and swallowing problems. Respiratory and pneumonia problems associated with ALS are the main reasons newly diagnosed patients are normally given just 3-5 years to live [5]. With approximately five thousand new diagnoses each year, there is a necessity for research in the topic [1]. Understanding the effects of ALS should aid in increasing the life expectancy of patients, with a long term goal of finding a cure.

Dr. Christopher Lee specializes in evaluating muscle ultrasound images of ALS patients in the Vanderbilt University Medical Center’s Neurology Department. He uses the information gathered to determine early clinical and electrodiagnostic features of critical illness polyneuropathy and myopathy. Dr. Lee’s current studies involve using the effect of atrophy in the muscles to gauge the progressed state of the disease in individual patients. The volume and shape of easily identifiable muscles like the bicep are measured over periods of months to observe differences brought on by the ALS disease.

Currently, the optimal method for measuring muscle volume is using Magnetic Resonance Imaging (MRI) [6]. This technique is the gold standard in image quality, but there are several factors that make MRI a difficult method for imaging ALS patients.

One issue with MRI is the high acquisition and operational costs of the machine. Typical MR machines run at approximately $3 million, a steep price for specialized clinics to be expected to fund. Operational costs vary depending on the size and strength of the magnet, but are usually high to cover the cooling process. A cheaper alternative would allow more patients to be seen and larger research projects to be conducted each year.

A second difficulty for ALS research conducted on MRI equipment is the relatively long acquisition times of approximately 30 minutes. Patients with ALS are prone to having uncontrollable muscle contractions or spasms that would reduce image quality if occurring during the acquisition time [7]. An imaging technique that uses a reduced time necessary for acquisition would benefit the image quality and patient stress.

A similar set of problems facing the use of MRI to image ALS patients is the immobile nature of the machine and its round bore. It can become difficult for people with ALS to move around and change positions in later stages of the disease as they near paralysis [7]. The lack of mobility in the patient means a portable imaging technique would better serve to ease the task of participating in research for the patient. Most MRI machines use round bores that require patients to lie prostrate for the full image acquisition time. Again, this is an issue for those patients whose disease has progressed to local or full paralysis and cannot easily lie down in the machine. There are MR designs with open bores, but they too require the patient to be flat in the machine. A technique for measuring muscle volume that does not include having the patient become uncomfortable due to movement or position would enhance the effectiveness of the research.

Overall, there is a need for a low cost, mobile, and fast method for accurate muscle volume measurements that accommodates the needs and restrictions of ALS patients. Ultrasound (US) is an alternative imaging technique and has been shown to give accurate muscle volume measurements, compared to MRI [8]. The machine used in this design, the Sonosite Titan, is much cheaper than an MRI machine at the relatively low cost of $17,000 [9]. The Titan machine is mobile, with a portable docking station, and is compatible with a laptop for data collection and analysis [10]. Ultrasound imaging is much faster than MRI, with imaging times remaining less than a minute, as reported by Dr. Lee.

The Nintendo Wii system uses infrared light and accelerometers to track the gaming controllers. Our design uses these qualities of the system to attempt to track the ultrasound probe as it acquires muscle images. The goal of our project is to integrate the ultrasound imaging method with tracking techniques from Nintendo’s Wii system to obtain accurate skeletal muscle volume measurements.

METHODOLOGY

System 1: WiiMote Tracking Four Infrared LEDs

Two WiiMotes were attached to individual goniometers using Velcro, and placed on a stable flat surface such that the centers of the WiiMote cameras were at a known orientation (55°), and distance apart (36”). Each WiiMote has a viewing angle of 41° and 31° in the horizontal and vertical axes, respectively [15]. The WiiMotes were paired to the computer via a USB bluetooth adapter. The infrared camera, accelerometer, and button status information were all parsed by the WiiLab MATLAB program [16].

The four 935nm infrared LEDs were mounted onto the circuit in a square arrangement as diagramed by the schematic in Figure 2, such that each side of the square was 1 inch. The LEDs have an emission angle of 90°. The prototype for the LED tracking module is shown in Figure 3. The module was then attached to the flat surface of the US probe with the LEDs being closer to the transducer. While testing, the prototype was oriented to face toward the WiiMotes.

Executing the bothwiimotes() function within WiiLab ran the LED triangulation and orientation software and initialized the US video feed, which was connected to the laptop via a S-video-to-USB converter. The triangulation software is based on the trigonometry of Figure 4.

Each WiiMote can simultaneously track up to 4 LEDs so the position of each LED was triangulated independently. Because the initial LED assignment order to the WiiMote was dependent on the order in which the WiiMote detected the LEDs it was essentially a random assignment each time the program was initialized. To institute some consistency in the assignment and correspondence between the WiiMotes, an LED assignment code was written to align the WiiMote LED registration with the LED location in space.

Knowing the WiiMote LED registration and the LED location in 3D, the orientation of the prototype and therefore the probe could be determined. Projecting through the xy, xz, and yz planes, the angle between 2 LEDs spatial location provided the orientation of the prototype with regards to each respective plane.

System 2: WiiMotionPlus and Wiimote Linear Accelerometer System

The second system we proposed employed the 3-axis linear accelerometers from the Wiimote used in conjunction with the 3-axis rate gyroscopic accelerometer data from the WiiMotionPlus to, in theory, provide the full six degrees of freedom needed to accurately track the US probe in 3-D space. The Wiimotes ADXL-330 3-axis accelerometer is used by the Wii system to estimate the yaw, pitch and roll angle of the Wiimote in relation to the Wii system’s infrared source (Figure 6). The accelerometer data was fed into MATLAB via the Wiimote’s Bluetooth transmitter. To obtain linear position data in 3-D space, the Wiimote accelerometer incoming data had to be integrated twice using numerical integration techniques in MATLAB. The Wii-Motion+ uses the InvenSense IDG-600 two axis rate gyroscopic accelerometer to provide the rates of change in yaw, pitch and roll.

The incoming data was read into MATLAB via an Arduino prototype board with a USB connection (Figure 7). To obtain yaw, pitch and roll angle from the Wii-Motion+ the incoming data had to be properly scaled and integrated using numeric integration techniques in MATLAB. The scaling was determined by repeatedly moving the WiiMotionPlus a known angle on each axis of rotation and scaling the incoming data appropriately. Both the Wiimote and WiiMotionPlus were securely attached to the US probe. Due to a computational error this device was not tested in full.

System 3: WiiMotionPlus and LED Tracking

The final system we designed employed the LED tracking system to provide linear position in 3-D space and the rate gyroscopic accelerometer data from the Wii-Motion+ to provide angular position. Two Wiimotes were used to track the IR array. The IR LED array and WiiMotionPlus were attached to the face of the US probe.

Phantom

To validate our method of measuring volume, we designed a phantom to be used by both the Sonosite Titan and a 4.7T small animal MRI machine. The phantom consisted mostly of a gelatin and psyllium concoction that has been used in previous studies to represent tissue in US images [17]. A straw was used to create an air bubble in the gel to provide an easily identifiable volume of space to be measured by the two techniques.

3-D Reconstruction Software

In order to re-create the ultrasound data, each image needed to be associated with its correct location and orientation. The use of the synchronized data acquisition technique makes this simple. Every slice was numbered according to order during data acquisition, as well as every data point and angle. In order to translate the data so that it represented the scanned area, an empty three dimensional matrix needed to be created that represented the size of the arm. The Sonosite Titan images were created using approximately 82 pixels per centimeter. Initially, the program asked for the dimensions, length, width, and depth of the arm in centimeters. From here the arm dimensions were multiplied by the 82 pixels per centimeter to give the appropriate number of pixels to represent the arm without distortion. In addition to the empty patient matrix, an empty temporary matrix and an empty count matrix of equivalent dimensions were created. Since the system returned the position data as a relative number, the data was normalized with the arm dimensions. After all the data had been appropriately scaled, one slice at a time was stored in the temporary matrix according to its x and y location. The top center point of each slice represented the actual location denoted by x and y. After being translated, the slice was then rotated according to the yaw, pitch, and roll angle. The program returned 3 matrices of the same size as the slice, but instead of holding intensity values, each held the new x, y, or z location. Given these new locations, the original intensity value was stored in the patient matrix. Every time a data point was stored in a particular voxel, it was noted in the count matrix. After the completion of the program, the patient matrix was divided by the count matrix to give an average intensity value at every location.

RESULTS

LED System

Following tracking, MATLAB was used to create a three dimensional plot of location. This plot included the location for each LED and appeared as though the tracking was working. Figure 8 shows the x, y, and z data returned from an ultrasound scan of the phantom. The data included a lot of noise so it was filtered using a 13 point low-pass digital filter. The scan returned 149 data points. As seen below the y data appears as if correct, because the scan remained still in this dimension. When looking at the x data however it appears incorrect because the motion involved a gradual decline along the x dimension. The z data look approximately correct however seem to still have a decent amount of noise. The z motion of the probe started at an initial location moved towards the IR detector, and then returned along the same path. When observing the angle data, the roll and pitch angles appear correct because they remain within 20 degrees of motion. The yaw however reaches angles of -80 degrees, this is obviously incorrect because the probe was never oriented at large angles.

In order to test the accuracy of the LED tracking system the device was moved along a fixed path. The data returned was then compared to the known data. This error is shown in Figure 9. The error is consistent with the data and observations from the ultrasound scan. Surprisingly, the greatest error shown is in the y data. This could be a result of the non-uniformity between the two IR cameras. A difference here could cause the Y position to remain constant when it is actually changing. The x error follows what was observed from the US scan. The z error is within range for the design specifications. Unfortunately, because the x and y error is greater than that of the ultrasound system the tracking system fails to meet the standards needed to reproduce accurate images.