Borrowing ideas from speech recognition research,
Johns Hopkins computer scientists are building mathematical
models to represent the safest and most effective ways to
perform surgery, including such tasks as suturing,
dissecting and joining tissue.
The team's long-term goal is to develop an objective
way of evaluating a surgeon's work and to help doctors
improve their operating room skills. Ultimately, the
research also could enable robotic surgical tools to
perform with greater precision.
The project, supported by a three-year National
Science Foundation grant, has yielded promising early
results in modeling suturing work. The researchers
performed the suturing with the help of a robotic surgical
device, which recorded the movements and made them
available for computer analysis.
"Surgery is a skilled activity, and it has a structure
that can be taught and acquired," said
Gregory D. Hager,
a professor of computer
science in the Whiting School of Engineering and
principal investigator on the project. "We can think of
that structure as 'the language of surgery.' To develop
mathematical models for this language, we're borrowing
techniques from speech recognition technology and applying
them to motion recognition and skills assessment."
Complicated surgical tasks, Hager said, unfold in a
series of steps that resemble the way that words, sentences
and paragraphs are used to convey language. "In speech
recognition research, we break these down to their most
basic sounds, called phonemes," he said. "Following that
example, our team wants to break surgical procedures down
to simple gestures that can be represented mathematically
by computer software."
With that information in hand, the computer scientists
hope to be able to recognize when a surgical task is being
performed well and also to identify which movements can
lead to operating room problems. Just as a speech
recognition program might call attention to poor
pronunciation or improper syntax, the system being
developed by Hager's team might identify surgical movements
that are imprecise or too time-consuming.
But to get to that point, computers first must become
fluent in the language of surgery. This will require
computers to absorb data concerning the best ways to
complete surgical tasks. As a first step, the researchers
have begun collecting data recorded by Intuitive Surgical's
da Vinci Surgical System. This robotic system allows a
surgeon, seated at a computer workstation, to guide tools
to perform minimally invasive procedures involving the
heart, the prostate and other organs. Although only a tiny
fraction of hospital operations involve the da Vinci, the
device's value to Hager's team is that all of the robot's
surgical movements can be digitally recorded and
processed.
In a paper presented in October 2005 at the Medical
Image Computing and Computer-Assisted Intervention
Conference, Hager's team announced that it had developed a
way to use data from the da Vinci to mathematically model
surgical tasks such as suturing, a key first step in
deciphering the language of surgery. The lead author, Johns
Hopkins graduate student Henry C. Lin, received the
conference award for best student paper.
"Now, we're acquiring enough data to go from 'words'
to 'sentences,'" said Hager, who is deputy director of the
Johns Hopkins-based Center
for Computer-Integrated Surgical Systems and
Technology. "One of our goals for the next few years is
to develop a large vocabulary that we can use to represent
the motions in surgical tasks."
The team also hopes to incorporate video data from the
da Vinci and possibly from minimally invasive procedures
performed directly by surgeons. In such operations,
surgeons insert instruments and a tiny camera into small
incisions to complete a medical procedure. The video data
from the camera could contribute to the team's efforts to
identify effective surgical methods.
Hager's Johns Hopkins collaborators include David D.
Yuh, a cardiac
surgeon from the School of Medicine. "It is fascinating
to break down the surgical skills we take for granted into
their fundamental components," Yuh said. "Hopefully, a
better understanding of how we learn to operate will help
more efficiently train future surgeons. With the
significantly reduced number of hours surgical residents
are permitted to be in the hospital, surgical training
programs need to streamline their training methods now more
than ever. This research work represents a strong effort
toward this."
Hager's other collaborators include Lin; Sanjeev
Khudanpur, an assistant professor of
electrical and computer
engineering; and Izhak Shafran, who was a postdoctoral
fellow affiliated with the university's
Center for Language
and Speech Processing and is now an assistant professor
at the Oregon Graduate Institute.
Hager cautioned that the project is not intended to
produce a "Big Brother" system that would critique a
surgeon's every move. "We're trying to find ways to help
them become better at what they do," he said. "It's not a
new idea. In sports and dance, people are studying the
mechanics of movement to see what produces the best
possible performance. By understanding the underlying
structures, we can become better at what we do. I think
surgery's no different."