Visual Servoing for Beating Heart Surgery - Visual Servoing via Advanced Numerical Methods - page 89

So there is a real need to reduce the residual motion of the AOI. This can be
obtained by canceling the relative motion between the AOI and the operation table
or by canceling the relative motion between the instrument and the heart. Both ap-
proaches have been investigated in the literature. They will be detailed in Section
6.2. Most of these techniques involving a robotic device use visual servoing. Indeed,
in a minimally invasive context, the endoscopic video of the AOI is always avail-
able. So it is relevant to use it as a feedback in a visual loop to regulate the relative
motion AOI-table or AOI-instrument towards zero.
in vivo images are known to be very difficult to process: specularities due to the
bright endoscopic light, large motions due to the high magnification factor and the
lack of structure make the feature extraction process especially difficult. Robustness
of the visual feedback is critical in order to meet the high safety standards for med-
ical devices. Redundancy of sensors can improve the robustness: fusing the visual
information with other signals can help maintaining a continuity of the feedback
during occlusions for example.
The analysis of the heart motion reveals that its motion has some characteristic
features [41]. This information can be used to make a predictive model of the AOI
future displacements. A good motion prediction can drastically improve the preci-
sion of the visual servo loop thanks to anticipated control signals [19]. A description
of the most recent prediction algorithms for local myocardium motion is given in
Section 6.3.
The surface of the heart is deformable with few landmarks: only the coronary ar-
teries network is visible. The vision algorithm should cope with this relatively poor
visual information. Tracking algorithm based on novel pattern matching techniques
have demonstrated their efficiency on live in vivo video of the myocardium. This
point will be discussed in details in Section 6.4.
Robust vision and efficient motion prediction are not the only keys to achieve effi-
cient visual servoing on the beating heart. The third key is dynamic control. Indeed,
heart motion dynamics are fast [13]. Low frequency visual servoing is unable to
cope with the sharp accelerations observed on the surface of the myocardium [13].
Furthermore, medical robots are lightweight and thus prone to flexibilities. Kine-
matic visual servoing [17] neglects these effects and so cannot achieve the optimal
bandwidth required by the task. In Section 6.5 the synthesis of advanced control
laws for efficient dynamic control is thus presented. Dynamic modeling of the vi-
sual loop is used in order to obtain the best performance. in vivo experimental results
for stabilization and tracking tasks are given.

6.2 Motion Compensation Approaches

In the context of TECAB, the surgeon's hand should simultaneously reach high
dexterity, to perform the anastomotic suture, and high dynamics to perform
surgery on an organ that exhibits large displacements with large accelerations. The
characteristics of the heart motion have been studied in details in the robotics com-
munity. On pigs, which heart has strong physiological similarities with the human