Giving Voice to New Technology
In his computer science laboratory on The George Washington University campus, James Hahn, Ph.D., holds what he calls a magic wand — a slender, black piece of plastic about eight inches long. Hahn places the wand near an ersatz patient — really just a hunk of white plastic that replicates a real human neck — and waves it. As he does so, a nearby computer monitor displays a cutaway view of the internal structure of the plastic block, including the simulated cartilage, windpipe, and vocal cords. Hahn flicks the wand to the left, and the image on the monitor tracks the move, acting as a digital window into the hidden anatomy humans need to talk.
“You can see the patient from any vantage point and even take a look from the inside,” says Hahn, chair of the Department of Computer Science and director of the Institute for Biomedical Engineering.
While the patient may be artificial, the need for such X-ray vision during voice-restoring surgery is very real. And no one knows this better than Steven Bielamowicz, M.D., professor of Surgery and an expert throat surgeon at GW. In collaboration with Hahn, Bielamowicz is the principal investigator on an interdisciplinary research project between the School of Medicine and Health Sciences and the School of Engineering and Applied Science. The project aims to develop computer-based tools to improve a rehabilitative surgical procedure for patients suffering from voice disorders caused by vocal cord damage or paralysis.
About one percent of adults develop some kind of grave vocal problem during their lifetimes, says Bielamowicz. Among the most serious vocal problems are those caused by injuries to the vocal cords or to the nerves that feed them, injuries that diminish an individual’s speaking ability — and sense of identity. Intubation during surgery, certain viruses, strokes, spinal surgery, tumors, or other trauma can all lead to the kind of damage that significantly restricts or even paralyzes one of the two vocal cords. When this happens, a person’s voice becomes hoarse and breathy, and the person struggles just to get out a few words. “There’s a great sense of effort and discomfort, and the quality of the voice deteriorates tremendously,” Bielamowicz says.
Fortunately, Bielamowicz enjoys vast experience performing a type of voice-restoring surgery called medialization laryngoplasty, which can remedy the problem. During the surgery, Bielamowicz makes an incision in the front of the throat and implants a thin sliver of plastic into the vocal cord. The implant supports the weakened or paralyzed vocal cord, pushing that cord into an ideal position for the opposing, healthy vocal cord to vibrate against to make sound.
Shaping and placing the implant are vital for a successful operation, but both tasks are tricky, as the anatomical details vary from patient to patient. “The surgery in its current form is very artful,” says Bielamowicz. “It’s only through a surgeon’s experience and his detailed understanding of vocal cord anatomy and function that he can restore vocal cord vibration and, thus, voice production.”
As the surgery gets under way, Bielamowicz has to track a wealth of information. He needs to characterize the exact type of abnormality that the patient has. “Currently we use very imprecise evaluation tools, when we look at the laryngeal anatomy and try to imagine how the system would work best,” Bielamowicz explains. Throughout the procedure the surgeon repeatedly looks down at the patient’s throat, then up at a monitor displaying data taken from an endoscope placed through the patient’s nose. The endoscopic images reveal the inner anatomy of the throat, showing Bielamowicz the position of the vocal cord. But the endoscopic image floats above the patient on a monitor — and the surgeon is left trying to imagine where, in the actual patient, the features seen in the endoscopic image are actually located. While the endoscopic image can serve as a guide, it’s not until the surgeon makes his incision that he knows exactly where his instruments are going to land in the anatomy.
And in vocal surgery, precision is everything. Placing the implant just a few millimeters too high or too low will render the device useless, and the patient’s voice won’t improve. In fact, despite Bielamowicz’s expertise — he has performed more than 1,000 medialization laryngoplasties over the past 15 years — about 20 percent of his patients need to make return trips to the operating room for minor adjustments of the implants. “That’s probably the most frustrating issue for the patients,” says Bielamowicz, who has operated on opera singers, famous radio hosts, and plenty of non-celebrities, too. “They come back for revision of the surgery, and we put in a larger implant or we shape the implant differently. If we had more information, I believe the need for these secondary surgeries would decrease significantly.”
Enter the biomechanical system, a surgical equivalent to X-ray vision that Hahn is developing. In 2005, he joined forces with Bielamowicz and engineering scientist Rajat Mittal, Ph.D., formerly of The George Washington University and now at Johns Hopkins University, to provide the extra information the surgeon needs to improve the success rate of vocal cord surgery. The team won a five-year, $2.8-million grant from the National Institutes of Health to develop the system. So far the project has progressed so steadily that the team has applied for a renewal of an NIH grant to continue its work.
The project has two main components. One is modeling the biomechanical systems for voice production through a computerized evaluation of airflow and tissue interaction. The second component is the development of a surgical image guidance system. The first will allow Bielamowicz to better plan his surgeries, while the second will direct him during the operation.
Mittal is developing the first component, an airflow simulator much like those used to design aircraft, to optimize the shape and placement of the plastic implant. During surgical planning, the patient will undergo a CT scan that builds a digital three-dimensional image of the throat from the top down. A second CT scan, taken as the patient speaks, provides a glimpse of how the compromised vocal cord is malfunctioning. “We then use the healthy vocal cord as a guide to see how the weakened cord should be moving,” Bielamowicz says.
Next, Mittal’s computer software simulates airflow through the windpipe during speech. The surgeon can then experiment with various virtual implants, moving different sizes and shapes around to restore normal vocal cord vibration and airflow.
These fluid dynamics simulations require a lot of number crunching — the software Mittal is developing renders the vocal cord as 300,000 data points that interact with many hundreds of thousands of moving air particles. Currently it takes about a day to crunch all the data to run the simulation, and the software needs more refining to be fully functional, says Hahn.
After the surgeon has settled on the perfect implant, it needs to be placed. This is when the second component of the project comes into play. The system Hahn is developing will provide Bielamowicz a real-time cutaway view of the interior of the throat during surgery. “This kind of image guidance will allow the surgeon to know exactly where he is with respect to the patient’s anatomy,” says Hahn. “In some sense, we’re giving the surgeon X-ray vision.”
The CT scan of the patient’s throat serves as the underlying basis of that vision. Inside a computer, Hahn’s software then layers two more images on top of the CT data. One image arrives from an endoscope, a small camera placed through the nose and into the throat, while the second image is sent from a stereoscopic camera placed above the patient. The software aligns the layers so that the endoscope image, which displays the inside of the throat in real time, perfectly aligns with the underlying cartilage and bone that the CT scan revealed (see caption, page 20). Likewise, the images from the camera are painted on top of the other two layers, and the entire “virtual throat” is displayed on a large monitor in the operating room.
By wiggling the endoscope’s magic wand, the surgeon can “fly” through the patient’s anatomy, burrowing beneath the skin and muscle, past bone and cartilage to the windpipe and the vocal cord. The wand — which is constantly tracked by the software — helps the surgeon pinpoint the exact spot on the voice box to make his incision and insert an implant.
“The goal is to provide a fusion of visual information,” says Bielamowicz. “Right now, it’s a lot to juggle in my mind. During surgery, I spend a lot of time looking up at the endoscopic image, then looking back down at the patient, then looking up again. If we can simplify that process, errors will decrease.”
Hahn and Bielamowicz have tested the system on cadavers, and they say it’s almost ready for prime time. One hurdle remains: As the surgeon makes his cuts and places the implant, the surrounding cartilage changes shape, throwing the three layers out of alignment. “Cartilage deformation is a real computational challenge,” says Hahn, who has enlisted several graduate students to help him find a software solution.
Once that problem is resolved, the team plans to test the system starting next year.
Bielamowicz predicts that, once operational, the system will dramatically reduce the need for repeat procedures while offering all patients better outcomes — and stronger voices.
“The ability to talk and to communicate is critical to our well-being,” says Bielamowicz. “It’s not just about singers or talk show hosts; it’s people who coach, people who are in sales and marketing, educators, lawyers, doctors. And even if you restore somebody’s voice, and it’s a functional voice but it doesn’t sound like them, that is unsatisfactory to the patient. They just want their voice back.”