Simulating Surgery in 3D for New Medical Device Design
Virtual Validation of In-Office Robotic Surgery
Like calluses that form on the hand, voice overuse can result in hard, noncancerous growths on the vocal folds. Removing these nodules promptly to preserve the vocal folds’ function is vital for actors and singers who rely on their voices. The traditional route to remove benign nodules, as well as some rare cancerous tumors, has been surgery in a hospital operating room.
In recent years, laser removal of such lesions in doctors’ offices has gained popularity. The procedure involves zapping vocal fold growths with laser pulses to destroy or shrink them. It can be a quick outpatient surgery that does not require general anesthesia.
However, the endoscopic device technology used for today’s procedures has limitations. “It works. This is the good news,” says Loris Fichera, a professor of robotics engineering at Worcester Polytechnic Institute (WPI). “The bad news is that it can be challenging to reach certain parts of the voice box because the instruments are very limited in how they direct light.”
New optical fiber designs and endoscope configurations should overcome this limitation. Fichera and his colleagues have used MATLAB® to simulate how fibers with angled tips, instead of the flat ones used today, perform best in high-definition 3D models of the human voice box, or larynx. By simulating the medical devices, they hope to reduce the time and cost needed to make physical prototypes that must be tested first on cadavers and then in the clinic. The simulations could provide validation during the development of more effective optical fibers and endoscopic instruments in the future.
a) The blue cone represents the view of the endoscopic camera, which includes a view of the tumor. The red dashed line represents the laser fiber line of sight. b) A tumor located underneath the vocal fold is less accessible to the forward-facing laser fiber. (Image credit: Carla Dipasquale Illustration, by permission of Loris Fichera)
“These devices cost a lot of money and time to manufacture,” Fichera says. “Using simulation with MATLAB, we can quickly evaluate dozens of different designs without fabricating them. We determine which configuration is promising and is the one we should prototype. This shortens prototyping time, making it much more cost-effective. Each cycle of design and prototyping now only takes between three weeks to two months, depending on the complexity of the design.”
Virtual Voice Box
Clinic-based larynx surgery involves passing a thin, flexible endoscope carrying a camera and an optical fiber into the throat. Today’s optical fibers have a flat end and emit light forward toward the target tissue, so they cannot reach the undersurface or edges of the vocal folds that are not in direct line of sight.
“With Medical Imaging Toolbox, we can load the entire data set and create the three-dimensional rendering with just a few clicks. Having this functionality and the capability to export data is important: It means we don’t start from scratch for each new design. We can rely on something we know works that is standard. This saves weeks for each new design.”
Recent research has suggested that tapered fiber tips that direct light at different angles could allow physicians to treat the hard-to-reach spaces of the larynx. Fichera partnered with medical professionals at Harvard Medical School and Brigham and Women’s Hospital in Boston to assess the effectiveness of such angled fibers. They conducted a simulation-based study comparing different types of laser fibers: a traditional forward-facing fiber and three side-firing fibers that emit light at angles of 45°, 70°, and 90°, respectively.
The WPI team created seven different 3D anatomical models of the human larynx by compiling images from microcomputed tomography (microCT) scans of larynx specimens and processing them with Image Processing Toolbox™.
The CT scans are a sequence of high-resolution grayscale 2D images in which each pixel conveys the density of the tissue at a location. Each image represents a slice of tissue that is a few micrometers apart from the slice in the following image. “It’s like having pictures taken at different heights,” Fichera says. “We process these images to reconstruct the three-dimensional shape, and the output of this is a stereolithography, or STL, file, the same format normally used for 3D printing.”
The 90° side-firing fiber can access previously unreachable areas of the larynx. (Image credit: Loris Fichera)
Commercial radiology software can reconstruct 3D shapes from medical images. Still, the software is challenging to work with and often cannot export data in the STL format that Fichera and his colleagues require. The team spent a week writing code to convert the CT scans into the 3D larynx rendering.
“With Medical Imaging Toolbox, we can load the entire data set and create the three-dimensional rendering with just a few clicks. Having this functionality and the capability to export data is important: It means we don’t start from scratch for each new design. We can rely on something we know works that is standard. This saves weeks for each new design,” says Fichera.
Fichera’s graduate students also use Medical Imaging Toolbox™ for his course on surgical robots. In the past, students were tasked with practicing three-dimensional reconstruction using radiology software. “There’s no need for them to use that software anymore,” Fichera says. “I ask them to use MATLAB for this.”
Simulating Surgery
After creating the virtual larynx models, the WPI team used open-source MATLAB code from File Exchange to create a model of a commonly used commercial endoscope. The endoscope has the tip of the camera over the right vocal fold and the tip of the laser fiber over the left. It moves with three degrees of freedom.
“With Parallel Computing Toolbox, the simulation takes about seven hours to run. Without parallel computing, it would take much longer, probably days.”
The team created a program to simulate the insertion and motion of the endoscope in the larynx using a motion planning algorithm. The researchers deployed the virtual endoscopes with the different optical fibers into the seven larynx models. The program produced a point cloud representing all the locations in the larynx that the instruments can reach.
Next, they used ray casting—a computational technique used widely by game developers to simulate illumination in a 3D virtual scene—to mimic the laser beam emerging from the tip of the optical fiber. They used a ray-triangle intersection algorithm to detect the larynx tissue reached by the laser beam. “Essentially, we go through every one of these cloud points one by one and ask, ‘OK, if our laser fiber is here and we are applying light from this location, what tissue do we reach?’”
The endoscope model showing the fiber’s degrees of freedom. (Image credit: Loris Fichera)
Ray casting is computationally expensive, he says, so they used Parallel Computing Toolbox™ to speed things up. “Instead of going through thousands of viewpoints and projecting light rays one by one, we run ray casting in parallel from as many points as possible at once, and then we aggregate all the results. With Parallel Computing Toolbox, the simulation takes about seven hours to run. Without parallel computing, it would take much longer, probably days.”
They found that the 45°, 70°, and 90° angled fibers increased the total larynx tissue that could be accessed and treated by an average of 50%, 74%, and 84%, respectively.
Designing Next-Generation Endoscopes
In addition to providing evidence that side-firing fibers could improve robotic laryngeal surgery, the study revealed a fundamental limitation of today’s endoscopes. Surgery simulations showed a glaring gap in the point cloud near the right vocal fold. “The software was telling us that there was a part of the larynx that was really hard to get to,” says Fichera. “So, if a patient presents with a tumor here, you cannot treat it. This did not make any sense, and the very first thing we thought was we must have made a mistake.”
“Our idea is to replace the endoscope used today with our own version of this device. Whatever design we come up with, we can first do a simulation in MATLAB before we fabricate or even touch any physical instrument, see if it does what we think it will do, and then move to prototyping.”
Surgery simulations identified a gap in the point cloud near the right vocal fold that currently available endoscopes cannot access. (Image credit: Loris Fichera)
To their surprise, Fichera’s physician colleagues confirmed that they do, in fact, experience this lack of access over the right vocal fold. Returning to the computer program, Fichera’s team realized that this gap was due to the eccentric design of the endoscope, with the camera on one side and optical fiber on the other, and because it is difficult for surgeons to rotate the endoscope through 360° fully.
Fichera says that endoscopes used for laryngeal surgery today are based on the design of other scopes instead of being optimized for this specific procedure. The WPI team plans to use their study as the foundation for developing new endoscopic medical devices.
“We started this research believing all we have to do is develop a new fiber but keep using the same instrumentation,” he says. “But ultimately, I think what these results are also telling us is that we have to rethink the design of the endoscope as well.”
Fichera and his colleagues are planning to use their data to apply for grants and use the funding to develop an entirely new device. “Our idea is to replace the endoscope used today with our version of this device. Whatever design we come up with, we can first do a simulation in MATLAB before we fabricate or even touch any physical instrument, see if it does what we think it will do, and then move to prototyping.”
Read Other Stories
MEDICAL DEVICES
Machine Learning Aids Neurosurgeon During Brain Surgery
Improving Parkinson’s Surgery with AI and Signal Processing
MEDICAL DEVICES
Machine Learning Helps Fix Broken Hearts
Startup Brings Arrhythmia Treatment into the 21st Century
