A Digital Stereo Microscope Platform for Microsurgery Training - Technologies and Applications of Artificial Intelligence - page 296

setup to practice surgical movements are described. In section 4 the results of a new
surgical motion learning experiment are described along with the conclusions.

2 Related Work

Surgical motion is considered a relevant predictor for analyzing surgical performance
[9-17]. These studies [9-17] focus on parameters such as the economy of motion and
patency. The problem with these studies is that they perform actual procedures and
make assessment based on the economy of motion. They are disadvantageous due to
(i) the subjective bias of the surgeons, (ii) the limited number of repetitions deter-
mined by the availability of the surgical cases and (iii) the lack of repeatability of the
procedure.
Other surgical motion studies [18-21] decompose surgical motion into sub-tasks
that can be independently evaluated. Existing analysis of surgical motion data is li-
mited to the time of completion and number of movements. Some evaluations use
error rate also as a measure. Without a means of assessing the quality of the move-
ment, these measures cannot be trusted.
Another group of researchers have constructed Hidden Markov Model (HMM) [19,
22] using the surgical motion. Though this study can simulate motions based on the
model, it does not help in training surgeons in using the surgical visual aids to pro-
duce better motion. HMM results are also difficult to explain to surgeons. Another
recent attempt to classify and assess medical procedure training using virtual reality
training systems is reported [30,31]. The motion based evaluation system is applied to
classify the motion made on a virtual reality trainer of bone marrow extraction proce-
dure.
Hence there are no suitable surgical training system to practice surgical motion with
immediate feedback. There is a need for a system that can help create a standardized
surgical motions and evaluate its quality. The evaluation must be deeper than the
measurement of the time or the length of the movement.
Dexterous hand motion is a result of a cooperative controlled movement of a num-
ber of hand joints. The mechanical joint motions produce tremors [23-25] and its
frequency is governed by the inertia and elasticity of the joint. If a method can isolate
the signals corresponding to the tremors of these joints and the linear motion, then
that method can be used to derive the basis data for computing the quality of motion.
Surgical motion represents a complex time series data denoting the working of a
large number of human joints producing both desirable and undesirable motion. Cap-
turing the motion through electromagnetic sensors may add instrument noise to the
already complex data. The surgical motion will have involuntary motions [23, 27,28]
such as physiological tremor in addition to the voluntary motion.
Though the range of frequencies of the components of surgical motion are known,
they are also known to vary. Thus they are seen spread across the spectrum. A spec-
tral decomposition method requires a model of the frequencies in the surgical motion.
One such known frequency formulation [24] is given in (1)

d(t) = X sin ( ˉ t) + 0.01 X sin 10 ( ˉ t) (1)