modality-based navigation
TRANSCRIPT
Injury, Int. J. Care Injured (2004) 35, S-A24—S-A29
Introduction/principle
Modality based navigation (MBN) means interactive
tracking of instruments in a co-ordinate system
defined by an imaging modality [1]. MBN relies on
a registration of the tracking system, eg, an optical
digitizer, with the imaging modality, eg, CT, MR, or
a fluoroscope. During the registration process, a
transformation matrix between the two co-ordinate
systems of digitizer and imaging modality is calcu-
lated. After this step, the tracking system ‘knows’
where the images will be generated. This registra-
tion process is procedure-unrelated and typically
carried out by a technical person at regular time
intervals.
As soon as the images have been acquired dur-
ing an intervention, instruments with appropriate
passive or active markers can be navigated imme-
diately within the image volume without further
preparation, especially without registration [2].
This procedure works because there is an inherent
match between the object and image. No property
of the patient anatomy is needed or used. In fact,
it is possible to navigate in or around any object,
even in symmetrical forms like a sphere that would
be relatively hard to register. It is also possible to
navigate mobile and non-rigid objects, eg, brain or
liver tissue, as long as they do not deform during
navigation through gravitation or manipulation.
In contrast, ‘conventional’ or in this context ‘pa-
tient-based navigation’ (PBN) systems do not have or
make use of the knowledge of exactly where the im-
ages have been generated. They use corresponding
Modality-based navigation
Peter Messmer1, Thomas Gross4, Norbert Suhm3, Pietro Regazzoni4, Augustinus L. Jacob2, Rolf W. Huegli2
1 Division of Trauma Surgery, Department of Surgery, University Hospital Zurich, 8091 Zurich, Switzer-
land2 Department of Radiology, University Hospital Basel, 4031 Basel, Switzerland3 AO-Development Institute, 7270 Davos, Switzerland4 Trauma Unit, Department of Surgery, University Hospital Basel, 4031 Basel, Switzerland
KEYWORDS:
Modality-based naviga-
tion; CT-based naviga-
tion; registration-free
navigation; technology
integration; radiation
dose; clinical accura-
cy; 2-D/3-D imaging
Summary1 Modality-based navigation (MBN) means the interactive tracking
of instruments in a co-ordinate system defined by an imaging modality, eg, CT,
MR, or a fluoroscope. During the registration process, a transformation matrix
between the two co-ordinate systems of the digitizer and imaging modality is
calculated. Navigation can start immediately after collection of the images with-
out an intraprocedural registration process. Since the imaging modality belongs
to the OR or the intervention suite, image update can be performed at any time.
Following a step-by-step procedure with navigation and image update in a rea-
sonable sequence, the risk for a virtual-real mismatch is minimized.
For CT-MBN, we obtained a freehand absolute positioning accuracy of 1.9±1.1 mm
in vitro. The in vivo freehand absolute positioning accuracy in pelvic fracture
fixation was determined to be 3.1 mm (unpublished data).
From our point of view, modality-based navigation is an efficient and safe alter-
native tool for computer aided interventions.
1 Abstracts in German, French, Italian, Spanish, Japanese,
and Russian are printed at the end of this supplement.
0020–1383/$ — see front matter � 2004 Published by Elsevier Ltd.doi:10.1016/j.injury.2004.05.007
Modality-based navigation S-A25
geometric features like points, contours or surfaces
that can be identified both in the image volume and
in the patient anatomy to re-match both reference
systems that were identical at the time of imaging.
In analogy to the definition given for MBN, PBN means
interactive tracking of instruments in a co-ordinate
system defined by a sequence of images.
Each of the above-mentioned navigation methods
comprises a set of invariants:
• The geometrical relation between the active part
of the instruments used and their respective mark-
ers recognized by the digitizer are not allowed
to change during navigation. For example, if a
surgical drill is tracked and the guidepin clamped
in the drill-chuck deviates from the axis of the
drill, the position of the tip of the guidepin can-
not be predicted from the position of the markers
anymore. There will be a difference between the
display of the navigation system and the real situ-
ation, a virtual-real mismatch.
• In MBN, the spatial relation between the digitizer
and imaging modality must be constant or the
modality has to be tracked.
• In MBN, the object to be navigated may not move
or be moved significantly between imaging and
navigation, otherwise the inherent match be-
tween image volume and imaged patient anatomy
will be lost.
• In PBN, the object to be navigated may not move or
be moved significantly between registration and
navigation, otherwise the match between image
volume and imaged patient anatomy reestab-
lished by the registration process will be lost.
• There are two ways around that limitation: Either
the movement is tracked by a dynamic reference
base (DRB) or it can be described with sufficient
precision in mathematical terms, eg, the linear
movement of an examination table.
• If the object to be navigated is tracked by a DRB
during the procedure, the geometrical relation
between the reference, eg, a clamp on a spinous
process, and the anatomy referenced, eg, a ver-
tebra, may not change during navigation.
Another very important consideration is a pos-
sible alteration of the object, or the anatomy, by
the procedure, in which case imaging and patient
anatomy no longer agree. This corresponds to a
map after an alteration of the landscape. There
are two possibilities to cope with that situation:
Redraw the map from scratch, which corresponds
to an intra-operative acquisition of a new set of im-
ages, or deform the existing map until it matches
the new real scenery, which is equivalent to some
type of elastic matching of the imaging volume to
the altered patient anatomy.
Materials and methods/implementations
MBN is, in principle, adaptable to different medical
imaging modalities. The CT-variant was developed
and described in Basel [3]. The fluoroscopic variant
comes in 2-D and 3-D-versions using either a con-
ventional fluoroscope or a fan beam CT like the ISO
C 3-D. It is already described in the paper of Nolte
and colleagues in this issue of Injury and will not
be dealt with here. GE developed an interventional
MRI (double doughnut) where imaging planes can
be dynamically defined by a pointer that is tracked
by an MBN [4].
CT-variant
What we describe here is the implementation in
Basel that has been operational for more than ten
years. Some details might be solved differently in
other implementations.
Set-up and registration
In CT-MBN, a constant relation between the imaging
plane and the 3-D digitizer (Fig. 1) is postulated. (It
Fig. 1: CT suite with 3-D digitizer mounted at the ceiling, allowing a constant relation between imaging plane and digitizer.
S-A26 P. Messmer et al:
would, of course, be possible to equip the CT-gantry
with a reference base to achieve free mobility of the
digitizer versus the CT-gantry. In fact, we did initially
do just that. The biggest drawback of this approach is
that the optical digitizer has to ‘see’ both the gantry
reference and the instrument reference, which tend
to be fairly wide apart. Thus, the digitizer is used at
the margins of its useable operating volume, where
it is less precise.) In our suites, the digitizer (Fig. 2)
is fixed at the ceiling. During registration, a radio-
translucent panel with several light emitting diodes
(LEDs) is brought into the imaging plane (Fig. 3). The
position is verified by imaging with the highest reso-
lution in the z-direction available, usually 0.5–1 mm.
Then the position of the LEDs is digitized and the
transformation matrix calculated. This registration
is done by a technician, taking about one hour and
is usually necessary every 3–4 months.
Clinical procedure
At the beginning of the intervention, the patient is
placed carefully on the CT-table (Fig. 4) and normally
taped to a vacuum mattress for immobilization.
Imaging is planned to include the path from entry
to target but no more. As a minimal variant, that
may just be a single slice. Every image is taken at
the invariant position of the imaging plane that was
localized in the digitizer co-ordinates in the regis-
tration step. All images are tagged with the table
increment that was current while it was taken ac-
cording to the DICOM-standard [5]. The image stack
is automatically transferred via a network to the
navigation computer that also receives the position
data from the digitizer. There, simple planning with
one or several trajectories is carried out. For each
trajectory, a starting point, normally at skin level,
(‘entry’) as well as an end point (‘target’) is defined.
The entry is marked on the skin with the help of the
CT laser guiding light and a grid perpendicular to
the laser. This process allows marking of the area
to be draped and enables a later plausibility check.
(Another option could be to determine the entry
with a non-sterilized navigated pointer directly on
the patient before draping.)
For better access during the intervention, the
patient is then moved out of the gantry into the
operating position. The current z-translation is
manually fed into the navigation computer to com-
pensate for that motion. Alternatively, the table can
also be tracked with an additional reference. After
Fig. 2: Close-up shot of the 3-D digitizer mounted at the ceiling.
Fig. 3: Radio-lucent panel with light emitting diodes brought into the CT-gantry for the calibration process.
Fig. 4: Patient before sterile draping lying on a vacuum mattress. Sometimes additionally an extension device is applied in order to gain a correct reposition before com-puter navigated intervention.
Modality-based navigation S-A27
disinfection, draping, and connection of the instru-
ment references to the navigation system (Fig. 5),
a plausibility check is performed: The tip of the
guidepin is placed on one of the entry points and
rotated around the tip in a cone-like fashion. If the
tip is shown at the corresponding place by the navi-
gation system and stays there during the rotational
movement, the system is supposed to work correctly.
Then the navigation can start. The current direction
and position of the guidepin should be brought into
line with the planned target path. The display gives
a visual feedback in the three principal planes and
with the aid of an ‘artificial horizon’ where lateral
translation as well as direction and magnitude of
angular deviation from the planned path are shown
(Fig. 6).
At the discretion of the interventionist, depend-
ing, for example, on the presence of vital structures
near the target path, the patient can be brought
back to the gantry for individual control scans. These
scans can replace older versions at the same table
position in the existing image stack, which again al-
lows taking of only a minimal set of images.
Clinical accuracy
The clinical accuracy of a navigation system can be
defined in different ways:
• As the repetitive accuracy of a registration process.
• As visible consistency of readout and surgical
scene.
• As avoidance of complications, eg, perforation of
a vertebral pedicle.
With MBN, the absolute error of the navigation
system can be determined on control images that
offer an independent, external verification of the
position. Planning and result are directly compa-
rable.
This absolute clinical accuracy in MBN is composed
from a number of different constituents:
• The accuracy of the position measurements of
single LEDs as well as the structure defined by an
assembly of several LEDs in a reference base.
• The accuracy of the registration of the modality
as well as the instruments.
• The skill of the interventionalist and their ability
not to bend the navigated instrument.
• The immobilization of the whole patient and the
organ to be targeted.
• The resolution of the imaging protocol used to
determine the position of the instrument.
• For CT-MBN, we determined the repetitive ac-
curacy of a position measurement of a reference
base with a FlashPoint 5000 (Image Guided Tech-
nologies, Boulder, CO) optical digitizer to be in
the magnitude of 1 mm, depending on the edge
length of the reference base and on the angula-
tion between camera and digitizer [3]. Repetitive
accuracy of CT-registration was 1.3 mm. In vitro
freehand absolute positioning accuracy was meas-
ured to 1.9 ± 1.1 mm.
• The in vivo freehand absolute positioning accuracy
in pelvic fracture fixation was determined (3.1
mm, unpublished data).
Fig. 5: Navigation reference mounted on the accumula-tor-drill.
Fig. 6: Navigation monitor giving visual feedback in the axial, coronal and sagittal plane as well as an artificial horizon including information about lateral translation and magnitude of angular deviation from the planned path.
S-A28 P. Messmer et al:
Radiation dose
CT is a high radiation dose imaging modality. There-
fore, special care must be taken to reduce the usage
as much as possible. There are several considera-
tions to be taken into account:
• The volume scanned should be restricted to a
minimum. This is possible with MBN since it is not
necessary to scan a larger volume solely for the
purpose of subsequent registration.
• The number of control scans should be restricted
to a minimum. This is possible with MBN since
single slices can be inserted into the initial im-
age stack.
• The dose per slice should be restricted to a mini-
mum. We showed that it is possible to reduce the
tube current and hence the dose by a factor of ten
as compared to diagnostic imaging without losing
guidance safety because of the high contrast of
cortical bone (unpublished data).
• The dose for the personnel should be restricted to
a minimum, which can normally be achieved by
everybody leaving the room except the patient. If
the patient is anesthetized or has to be monitored
closely for other reasons, very low doses occur im-
mediately next to both sides of the gantry [6].
MR-variant
MBN-systems have been described, eg, in conjunc-
tion with the experimental open magnet from Gen-
eral Electrics known as the ‘double doughnut’ [4].
If an open MR is not available, MBN can be imple-
mented much the same way as the CT-variant, with
appropriate changes of the registration and naviga-
tion hardware. The general accuracy is expected
to be slightly worse than with CT due to spatial
distortions caused by magnetic field inhomogenies,
device image distortion, signal cancellation, and
susceptibility artifacts.
Fluoroscopy variant
The fluoroscopy variants of MBN, with both 2-D and
3-D imaging, are commercially available. They are
described in detail in the contribution by Nolte et
al in this issue of Injury.
Briefly, the navigation system localizes the fluoro-
scope, which is equipped with references, immedi-
ately prior to imaging. Since the relation between
the references and the imaging geometry has been
previously calibrated, the navigation system ‘knows’
the spatial position of the images acquired. When
the imaging is done, the imager can be put aside
to expose the surgical field. This is symmetrical to
the movement of the patient out of the gantry in
CT-MBN. If the object of interest is equipped with
a dynamic reference base, it can be moved during
the subsequent manipulation without losing naviga-
tion information. 2-D and 3-D variants differ in how
detailed their displayed anatomy is. The field of view
of current fluoroscopes is limited to about 20 cm in
2-D projection radiography and to a cube of 12 cm
edge length with fluoro-CT.
Discussion
MBN is a natural and powerful extension of an imag-
ing modality that is available in the intraprocedural
setting and whose geometric properties are known
with sufficient accuracy. In a registration process, the
spatial relation between a digitizer and the modality
has to be determined. MBN is functional immediately
after imaging and allows a seamless integration of
additional intraoperative control images.
MBN has to be compared to online imaging meth-
ods like fluoroscopy and CT-fluoroscopy on one hand,
and to ‘conventional’ 2-stage navigation systems
using image preserves and registration on the other
side.
Compared to online image guidance with fluoros-
copy or CT-fluoroscopy, MBN allows arbitrary inclina-
tions of the instruments to the imaging axis or plane.
There is unrestrained access to the patient since the
modality has been removed from the operating field
or vice versa. The dose to patient and personnel in
fluoroscopy strongly depends on the techniques (fo-
cus-object versus object-image intensifier distance,
coning down, pulsed fluoroscopy, shielding, and oth-
ers) and operator experience. With MBN, the dose
to the personnel is substantially reduced, in most
instances to zero, since everybody can leave the
room. The dose to the patient in MBN depends on
the factors outlined in “Radiation dose”. Procedural
safety of MBN is very high and comparable to online
imaging when control images are acquired at critical
points to detect a potential virtual-real mismatch.
MBN has several advantages compared to PBN:
Navigation can start immediately after collection
of the images without an intraprocedural registra-
tion process. It is a 1-stage procedure and less time
consuming than PBN. There is no need to image [7]
or surgically expose the anatomy only to identify
landmarks. PBN exists either with or without preop-
erative imaging such as CT-based or CT-free naviga-
tion. This preoperative imaging may lead to a high
radiation dose because of a large scan area.
Intraoperative image update guarantees the cor-
rect positioning of instruments. A final image check
Modality-based navigation S-A29
serves as total quality management: At the end of
the day, there is no doubt about whether a procedure
was morphologically successful or not.
The drawback of MBN is the necessity to have the
chosen imaging modality in the therapeutic suite,
which adds complexity. On the other hand, the need
for intraoperative imaging is being recognized in
more and more surgical disciplines, making imaging
modalities more available in surgical settings. The
problems of added complexity must in future be
solved by a higher level of technological integration.
As soon as a capable imaging is in the therapeutic
environment, MBN is an attractive option.
The question of which imaging modality is preferable
in which situations remains. In our view:
• 2-D fluoroscopic MBN is attractive in anatomically
simple situations, eg, in long bone injuries, where
soft tissue depiction is not required.
• 3-D fluoroscopic MBN shines in anatomically com-
plex situations, eg, in ankle or wrist fractures,
where the volume to be imaged is small and soft
tissue depiction is not required. 3-D imaging of
the spine may therefore also be an appropriate
indication.
• CT-MBN is perfect for anatomically complex situa-
tions where the volume to be imaged may be large
and soft tissue depiction is helpful, eg, in pelvic
or trunk trauma.
• MR-MBN is best in anatomically complex situations
where soft tissue depiction is paramount, as in
neurosurgery.
In conclusion, modality-based navigation is a
very efficient and safe tool for computer aided
interventions. Several imaging modalities may be
used, depending on the clinical situations and avail-
abilities.
References
1. Messmer P, Baumann B, Suhm N, et al. (2001) Navigation
systems for image-guided therapy: A review. Rofo Fortschr
Geb Rontgenstr Neuen Bildgeb Verfahr; 173: 777–84.
2. Russel Taylor SL, Burdea G, Moesges R (1996) Computer-Inte-
grated Surgery. MIT Press.
3. Jacob AL, Messmer P, Kaim A, et al. (2000) A whole-body
registration-free navigation system for image-guided surgery
and interventional radiology. Invest Radiol; 35: 279–88.
4. Nabavi A, Gering DT, Kacher DF, et al. (2003) Surgical naviga-
tion in the open MRI. Acta Neurochir Suppl; 85: 121–5.
5. Bidgood WD Jr, Horii SC, Prior FW, et al. (1997) Understanding
and using DICOM, the data interchange standard for biomedi-
cal imaging. J Med Inform Assoc [Am]; 4: 199–212.
6. Gebhard F, Kraus M, Schneider E, et al. (2003) Radiation
dosage in orthopedics–a comparison of computer-assisted
procedures. Unfallchirurg; 106: 492–497.
7. Schaeren S, Roth J, Dick W (2002) Effective in vivo radiation
dose with image reconstruction controlled pedicle instrumen-
tation vs. CT-based navigation. Orthopade; 31: 392–396.
Correspondence address:
PD Dr. Peter Messmer
Division of Trauma Surgery
Department of Surgery
University Hospital Zurich
8091 Zurich, Switzerland
E-Mail: [email protected]