Computer-Assisted Navigation on the Arrested
Heart during CABG Surgery
C. Gnahm1, C. Hartung1, R. Friedl2, M. Hoffmann3, K. Dietmayer1
1Institute of Measurement, Control and Microtechnology, University of Ulm
2Department of Cardiac Surgery, University Hospitals of Ulm and L¨ ubeck
3Department of Diagnostic Radiology, University Hospital of Ulm
Abstract. Coronary artery bypass grafting (CABG) is the standard
treatment for advanced coronary artery diseases (CAD). Optimal place-
ment of the bypass graft on the diseased vessel is very important. To as-
sist the surgeon in this matter, a computer assistance system has been de-
veloped facilitating navigation on the heart surface. After retrospective
validation on patient data sets yielded good results, the system was suc-
cessfully applied for in-vivo navigation on the front side of the heart dur-
ing several CABG procedures. Postoperative evaluation confirmed that
intraoperative navigation has been performed with adequate accuracy.
The underlying pathology of CAD is intraluminal narrowing of the coronary
arteries by wall plaque formations, severely limiting the blood flow to the my-
ocardium. During CABG surgery, the blood flow is rerouted around a stenosis
by a blood vessel graft. The position of the distal anastomosis on the diseased
vessel is of great importance. Optimal placement of the anastomosis is a difficult
task, especially as the target region may be covered with epicardial fat. To assist
the heart surgeon in optimally placing the bypass graft, a computer assistance
system facilitating navigation on the arrested heart during open heart CABG
has been developed and tested.
2Materials and methods
Nowadays, navigation systems are widely used in surgical disciplines such as
neurosurgery and orthopaedic surgery [1, 2]. Concerning cardiac interventions,
valuable research has been done in minimally invasive surgery . In open heart
surgery, the development of navigation systems is still at the beginning. The
Cardio-Pointer project is the first approach to develop a surgical assistance sys-
tem for open heart CABG with navigation at the beating  and the arrested
heart. The system concept comprises preoperative planning of the optimal po-
sition for the anastomosis as well as intraoperative navigation to that position.
Navigation is facilitated by registration of pre- and intraoperative data of the
patient’s heart. This is illustrated in fig.1 which also shows the intraoperative
setup and the surgeon using the Cardio-Pointer (trackable pointing device).
Computer-assisted navigation 263
2.1Preoperative data set and preoperative planning
A cardiac MSCT (multi slice computed tomography) showing stenoses of the
coronaries is recorded prior to surgery.
optimal distal anastomosis site is planned jointly by heart surgeon and diagnostic
radiologist and marked in the MSCT data. Apart from forming the basis of the
preoperative planning, the MSCT data are used as a basis for the registration
and intraoperative visualisation. For the registration, 3D positions of coronary
vessels and anatomical point landmarks potentially visible during surgery are
extracted. For the visualisation, a volume rendered map of the coronaries is
For each diseased target vessel, the
2.2 Intraoperative data set and registration
During surgery, the arrested heart is repositioned prior to grafting each bypass.
As it is intended to represent the operating situs, intraoperative data have to
be recorded after positioning the heart for subsequent bypass grafting. During
the short period of data recording, registration and navigation, movement of
heart or patient must be avoided. Intraoperative data are recorded with an
infrared optical tracking system. Using the trackable Cardio-Pointer, the surgeon
retraces visible parts of the vessels and points at anatomical point landmarks.
In most patients, only a few point landmarks and vessel paths are actually
visible on the heart surface due to epicardial fat. If the number of corresponding
point landmarks is smaller than 3, additional landmarks can be generated in
most cases. Then, all point landmarks are employed to coarsely match the
data sets. The matching is then refined using the recorded vessel paths. Further
improvement of the matching is achieved by a subsequent deformation correction.
A more detailed description of the registration mechanism and the possibility
of generating additional landmarks based on the vessel length was previously
described in .
intraoperative data for navigation; Cardio-Pointer on the heart surface.
From left: Intraoperative setup; Schematic view of the usage of pre- and
264 Gnahm et al.
Table 1. Overall errors and standard deviation for 16 vessels and 48 LMs
Mean RMS of point LMs (in mm)
RMS standard deviation of point LMs (in mm)
Mean RMS of vessel paths (in mm)
RMS standard deviation of vessel paths (in mm)
2.3 Image-guided navigation
Registration establishes a direct connection between the preoperative data in-
cluding the planned anastomosis site and the intraoperative data representing
the operating field. This allows for visualisation of the planned anastomosis site
together with the current position of the pointer in the patient’s coronary map.
A video screen is used for visualisation. The displayed pointer position is con-
tinuously updated while the surgeon moves the pointer along the target vessel.
Consequently, navigation to the preoperatively planned position is enabled.
The registration mechanism developed for the arrested heart was validated in
two phases: At first, intraoperative tracking data sets recorded during bypass
grafting procedures have been registered retrospectively with preoperative CT
data. The registration results obtained this way were very promising and justified
subsequent in-vivo testing of the registration mechanism. Thus, the intraoper-
ative modus operandi was altered to enable not only data recording but also
immediate registration and subsequent navigation on the arrested heart.
3.1 Evaluation of retrospective registration
Retrospective registration was performed on 16 data sets from patients under-
going CABG surgery. Intraoperative data were recorded directly prior to bypass
grafting after the heart was manually brought into a position exposing the target
vessel in the centre of the operating field. Although considerable deformations
of the heart were observed in each case, the registration mechanism for the ar-
rested heart was able to match all data sets with sufficient accuracy, i.e. a similar
accuracy is found for multiple measurements of the same anatomical landmark.
10 of the 16 data sets provided at least the minimally required number of 3 cor-
responding point landmarks, while 6 data sets didn’t provide enough landmarks
due to epicardial fat. For these data sets, 1 to 3 additional landmarks were gen-
erated on the basis of the unaltered vessel length. To evaluate the performance
of the registration mechanism, both the RMS (root mean square) fiducial regis-
tration error of all corresponding landmark (LM) points and the average distance
Computer-assisted navigation 265
of corresponding vessel paths were calculated after registration (Tab. 1). Good
results led to intraoperative live application of the system.
3.2 Intraoperative registration and navigation
Intraoperative navigation on target vessels on the front side of the heart has been
performed during 7 bypass grafting procedures. Both, the complete registration
mechanism and an alternative registration based on the vessel length  were
used for live application. The resulting accuracies indicated no preference for
either method. After successful registration, the surgeon performed navigation
with the pointer on the arrested heart guided by the surgical assistance sys-
tem. Target vessels on the front side of the heart were the LAD and its diagonal
branches. In the patient’s coronary map, the optimal position of the anastomosis
as preoperatively planned by the heart surgeon and radiologist was displayed.
After registration, the current position of the pointer was also displayed and
continuously updated while the heart surgeon moved the pointer until its cur-
rent position coincided with the planned anastomosis site (Fig. 2 (a)). This
position was marked with a clip right beside the vessel and then, the bypass was
grafted. Navigation on a target vessel including intraoperative data recording
and registration takes about 3-4 minutes. During this short period of time, nei-
ther the heart nor the operating table must be moved. In two cases, successful
registration was performed, but valid navigation was not possible due to unin-
tendedly induced movement. This is attributed to the learning curve.To evaluate
the navigation results, a postoperative MSCT of the patient was compared with
the preoperative MSCT data. In both data sets, the same distinctive point on the
target vessel path was used as reference point (Fig. 2b). The length of the vessel
centreline from this reference point to the position of planned and grafted anas-
tomosis respectively was determined. Comparison of both lengths shows how
accurately the bypass was grafted to the preoperatively planned position. In case
the position of the grafted anastomosis was not right beside the clip marking the
navigation result, the clip position in the postoperative MSCT was evaluated.
As a grafted anastomosis is several mm long, the centre of the anastomosis was
Fig.2. (a) Intraoperative navigation: planned anastomosis site (dark +) and current
pointer position (bright x) are displayed in the coronary map for one arbitrary moment
during navigation and the end of navigation.
preoperative and postoperative MSCTs with the distances from a bifurcation to the
planned (left) and grafted (right) anastomosis respectively along the vessel centreline.
(b) Evaluation of navigation result:
266 Gnahm et al.
Table 2. Navigation results
Target Distance from planned
vesselto grafted site (in mm)
Table 3. Target vessels and errors
Number of LAD bypass grafts
Number of bypass grafts on
diagonal branches of the LAD
Mean distance of planned and
grafted anastomosis (in mm)
Standard deviation of the
distances (in mm)
used for calculations. Measurement inaccuracies of up to ±2mm may occur. For
the 7 target vessels, the distance between planned and grafted anastomosis was
2.3 ± 1.6mm, confirming good accordance of planned positions and navigation
results (Tab. 2 and 3). This navigation accuracy actually reaches the limiting
measurement accuracy at the arrested heart which amounts to ±2.2mm.
A surgical navigation system for computer assisted CABG on the arrested heart
has been developed on patient data sets and its performance has been evaluated
during several intraoperative live applications. The system successfully enabled
intraoperative navigation on the surface of the heart during CABG surgery.
The heart surgeon efficiently performed navigation with the Cardio-Pointer on
the target vessels for bypass grafting. Postoperative evaluation shows that the
intraoperative navigation correctly identified the planned anastomosis site in
each case with sufficient accuracy. Navigation on all other target vessels for
bypass grafting has been performed as well, the results currently being evaluated.
The system developed within the Cardio-Pointer project is capable of matching
pre- and intraoperative data sets from the patient’s heart and provides precise
intraoperative navigation assistance to a preoperatively planned position.
1. Peters TM. Image-guided surgery: From x-rays to virtual reality. Comput Meth
Biomech Biomed Eng. 2000;4(1):27–57.
2. DiGioia AM, et al. Computer assisted navigation systems for hip and knee recon-
structive surgery. In: DETEC. ASME 2001; 2001.
3. Coste-Mani` ere E, et al. Optimal planning of robotically assisted heart surgery:
First results on the transfer precision in the operating room. Int J Robot Res.
4. Hartung C, et al. Image guidance for coronary artery bypass grafting. Int J CARS.
5. Gnahm C, et al. A registration approach for cardiac MSCT and optical tracking
data recorded at the ischaemic heart. Proc MIUA; 2008. p. 251–255.