Automatic Detection and Segmentation of Robot-Assisted Surgical Motions

1
Automatic Detection and Segmentation of
Robot-Assisted Surgical Motions

• THE RESULTS
• Linear discriminant analysis a robust tool for
  reducing and separating surgical motions into a
  space more conducive to gesture recognition.
• Achieved gt90 recognition rates on 15 datasets.
• In one test, we reduced 72 feature vectors into 3
  dimensions with 6 classes and still achieved
  nearly 90 recognition.

• THE PROBLEM
• Need to assess and quantify technical surgical
  skill objectively.
• Surgical training traditionally an interactive
  and slow process.
• Lack of fully documented surgical procedures that
  can be quickly mined or stored.

The result of LDA reduction with m6 and d3. The
expert surgeons motions (left) separate more
distinctly than the less experienced surgeons
(right).
A video frame of the suture task used for this
study.

• THE SOLUTION
• Recognize elementary motions that occur in a
  simplified surgical task.
• Robot motion analysis of users with varying
  daVinci (Intuitive Surgical) robot experience
  studied.
• Divided task into functional modules. Applied
  statistical methods, such as linear discriminant
  analysis (LDA) and probabilistic Bayes
  classifier.

Left, the results of grouping the motion
categories and varying the dimension of the
projected space. In the second column, the number
of unique integers indicates the number of motion
categories, and the position of the integer
indicates which motions belong to that category.
Right, the results of testing a recognition model
trained on both expert and intermediate surgeon.
Class was set to 12345566, with t10 and s2.
Note that both expert and intermediate data was
similarly recognized.

• PEOPLE INVOLVED
• Graduate Students Henry Lin, Todd E. Murphy
• Engineering Faculty Gregory D. Hager, Ph.D.,
  Izhak Shafran, Ph.D., Allison M. Okamura, Ph.D.
• Medical Faculty David D. Yuh, M.D.

Functional block diagram of the system used to
recognize elementary surgical motions in this
study.

• REFERENCES
• Henry Lin, et al., Automatic Detection and
  Segmentation of Robot-Assisted Surgical Motions,
  submitted to MICCAI, 2005.

• SUPPORTED BY
• NSF

A plot of the Cartesian positions of the left
master, identified coded by surgical gesture,
during performance of a suturing task. The left
plot is that of an expert surgeon while the right
is of a less experienced surgeon.
Engineering Research Center for Computer
Integrated Surgical Systems and Technology