Back to Annual Meeting ePosters

OBJECTIVE:
In order to synchronize robotic surgical instruments during totally endoscopic coronary bypass grafting with the heart motion the estimation of the position of the intervention point is required. In this study a model-based heart motion estimator is presented and validated in beating artificial heart model.
METHODS:
The heart wall geometry is approximated by several connected pulsating, elastic membranes subjected to deformation forces. The motion of the membranes is described by a physically-based state model. Additionally, a measurement model which maps the state to position measurements of the heart surface is defined. An estimator is used to predict the state of the motion model and update the model state by means of noisy measurements. For evaluation of the reconstruction accuracy three image sequences of the beating artificial heart without/with occurred occlusions are analyzed. The estimator is evaluated by tracking the landmarks on a moving artificial heart with a stereo camera system.
RESULTS:
The average absolute error in the image sequence without occurred occlusions ranged from 0.76-4.18 mm at the evaluation points (EVP=5). In the image sequence with simulated occlusions the absolute error ranged from 0.62-3.66 mm at the evaluation points. In the image sequence with a really occurred occlusion caused by a swab is ranged from 0.58-3.07 mm at the evaluation points. The image processing software tracks the landmarks reliable by short and long-term occlusions.
CONCLUSIONS:
The model-based stochastic approach allows a reliable estimation and reconstruction of the distributed heart motion based on spatial-discrete measurements of the heart surface.