Most people associate ultrasound technology with pregnancy and the little heartbeat on the monitor. A researcher at the University of Pittsburgh has a slightly different application in mind.
Nitin Sharma, assistant professor of mechanical engineering at Pitt, recently received more than $500,000 from the National Science Foundation to develop algorithms that could measure muscle function in patients with partial paralysis due to spinal cord injuries — just by looking at ultrasound images of affected areas.
By stimulating muscles electrically and using a sort of robotic leg brace called an exoskeleton, Sharma can already help patients with both total and partial lower-body paralysis walk a few steps. This exercise can help the patients with partial paralysis regain movement through repetition.
https://www.youtube.com/watch?v=e4suNzz5RxU
Sharma explained that in cases of partial paralysis (or incomplete spinal cord injury), the goal is to have the exoskeleton and/or electrical stimulation do only the minimum amount of work necessary — and determining that minimum requires measuring the person's remaining voluntary muscle function.
"If we don't measure it correctly, the robot could be over-compensating or it could be under-compensating and this becomes very critical," said Sharma. "If the robot under-compensates, it could lead to some instability or a fall. If the person is doing less work and the robot is doing more work ... it slows the learning process [for the muscles]."
Sharma said that currently, the best non-invasive strategy he has for taking these measurements is called surface electromyography, or EMG, which involves attaching electrodes to the legs to measure the electrical signals associated with muscle movement. But that technique has some drawbacks.
"It's listening to all of the muscles that are working, so it's very non-specific," said Sharma.
That becomes an issue when trying to measure the function of smaller muscles grouped very closely together, like the ones in the ankle.
Furthermore, Sharma said that electrical stimulation and EMG do not go well together.
"With EMG, we're listening at very low levels, while electrical stimulation is at much higher levels, so it creates some stimulation artifacts," said Sharma.
In other words, traces of the stimulation's electrical signal show up in the EMG measurements, creating inaccurate results. This is a problem because the idea is to have the robotic/stimulation assistance respond to feedback from the muscle function measurements in real time. Filtering out the artifacts would require the implementation of special signal-processing circuits in the equipment, which Sharma described as "a complicated job."
This is where the ultrasound imaging represents a potential advantage.
"The key thing is that you're looking directly at the muscle, you're looking at the muscles that you [actually] want to look at," said Sharma.
The first challenge for Sharma and colleagues is to actually develop the algorithms which can accurately read the ultrasound images to calculate the voluntary muscle contractions by the subject.
Sharma said they will start out by using an ultrasound machine similar to what you might find in a hospital to look at the ankles of patients. Once the algorithms have been properly developed, Sharma said the next step would be to try to implement wearable ultrasound technology into the existing exoskeletons.