192 / 236
registration errors were calculated using a standardized paired-
point algorithm. Custom navigation software, developed by our
group, provided a real-time 3D virtual view that was displayed
on a submonitor adjacent to the endoscopic monitor (Fig. 2a,
b).
7
Surface contours were displayed as a mesh and based on
intraoperative CBCT imaging (Fig. 2c). A traditional triplanar
representation of the updated CT images was placed on the sub-
monitor. The drill was tracked live, and its position was shown
on both the virtual and cross-sectional views (Fig. 2d). Auditory
feedback was provided through alarms when the tracked instru-
ment (drill) entered a contoured volume representing a critical
structure or a proximity alert zone.
Sixteen heads were dissected in total. Two heads were dis-
sected as a pilot by the investigators to establish initial labora-
tory setup and fine-tune alarm and visualization settings. The
seven subjects then performed an endoscopic transclival
approach on two heads each; one with the LIVE-IGS system and
one in a conventional manner (standard IGS involving a tracked
probe and preoperative CT) in random order. The extent of dissec-
tion was to each carotid artery laterally and to the dura posteri-
orly. Ablation extended from the pituitary fossa down to below
the level of the petrous segment of the carotid arteries.
The National Aeronautics and Space Administration Task
Load Index (NASA-TLX) questionnaire was administered mid-
way through the exercise and again at completion.
12
Open feed-
back was encouraged during the trial. After completing both
dissections, a semistructured interview was carried out and
recorded on video. Open responses on general use of the system
and each feature were acquired prior to more in-depth
questions on specific elements, their potential uses, and recom-
mendations. A seven-point Likert questionnaire pertaining to
aspects of the trial was also administered.
The recorded interviews were scrutinized by three investiga-
tors. Feedback was categorized, and comments on these features
were documented then communicated to an independent observer.
Agreement in regard to statements was reached, and these data
were entered into a consensus document in subcategories.
Statistical Analysis
Likert questionnaire responses are displayed as medians
and interquartile range (IQR). The Wilcoxon signed rank test
was used to compare the paired NASA-TLX responses, with
P
<
.05 deemed significant.
RESULTS
All seven participants completed the two clivus
ablation exercises. A significant amount of time was allo-
cated to become familiar with the navigation system and
to explore its capabilities. Fiducial registration errors
were consistent with current clinical practice (between 1
mm and 1.8 mm for all cases).
Initial Open Feedback
All subjects were generally impressed by the fea-
tures displayed, particularly the ability for real-time
Fig. 2. (a) Three-dimensional virtual view displaying wall-down view of critical structures behind a white mesh overlay of surface contours.
The tracked drill is shown. The virtual view updates in real time to provide a view perceptually matched to the endoscopic view. (b) Wet labo-
ratory setup showing the virtual view parallel to the endoscopic monitor in addition to the triplanar views. All images update dynamically as
the endoscope and drill are tracked by an image-guided surgery camera located above the display. (c) Cone-beam computed tomography
used to update imaging after the initial surgical approach and before clivus ablation. This allowed refinement of the display to show more
realistic surface contours. (d) Optical tracking registration markers placed on the head, endoscope, and drill. Orientation of markers was
adjusted to limit line-of-sight interference.
Laryngoscope 124: April 2014
Dixon et al.: Real-Time Navigation for Endoscopic Surgery
170