CAS in soft tissue surgery
The purpose of this (PhD thesis) project is to demonstrate the feasibility of updating the position and shape of the 3D virtual representation of a deformable object with measurements made on the real object. This work is part of the research carried out by the VRAI Group in the field of Computer-Assisted Surgery (CAS) where the goal is to superimpose virtual information to imaging modalities used routinely in the operating theater. This fusion of real and virtual data is called augmented reality. In the case of liver surgery and in particular for hepatic resections, the surgeon has to decide which functional part(s) of the liver should be removed in order to clear the tumor. The functional parts of the liver (Couinaud’s segments) are defined relative to the hepatic veins and the surgeon tries to situate himself relative to these vessels and locate the tumor by means of ultrasound imaging. The methods developed in this project aim to ease this localization procedure and guidance while enhancing the accuracy of the surgical operation.
Providing the surgeon with accurate visual aids such as 3D models is still an open problem in soft tissue surgery. The problem comes from the fact that pre-operatively generated models are frozen into the shape the organ had at the time of the acquisition of the scanner images. During surgery, the organ becomes deformed, caused by multiple factors such as surgical manipulation, decreasing vessel turgescence, respiratory and heart beating motions, and the organ no longer corresponds to the model. As a consequence, rigid models are not suitable for augmented reality with soft tissues and it is necessary to update pre-operative models during surgery to cope with organ deformation. However, in order to respect surgery time constraints a compromise between accuracy and speed must be achieved.
We are primarily focusing on localizing and updating the shape of the 3D pre-operative model of liver vasculature. This appears to be of major interest to the surgeon since all decisions taken are made according to the hepatic vessels. In deed, an appropriate margin around the tumor must be defined by taking care of not damaging the remnant vessels. As liver vasculature is dense then, measuring its location will enable an approximation of the deformation of the entire organ and define where the resection plane is positioned on the actual organ. The main steps of our method can be summarized as follows:
- Pre-operative generation of a mass-spring skeleton based on the 3D reconstruction of hepatic vessels. The skeleton will be used to deform the model according to the measurements.
- Intra-operative ultrasound imaging of the liver with a tracked US probe (2.5D). Images are segmented in order to extract the position of the vessels and generate a 3D cloud of points defining their center-line.
- Determination of the correspondence between the 3D cloud of data points and the model skeleton based on graph matching techniques.
- Update of the pose and shape of the model skeleton caused by attraction forces generated from the data points (non rigid registration).
|Mass-spring model of deformation. The masses and springs are located on the skeleton of the mesh of the liver vascular network. Mesh vertices are rigidly linked to the skeleton nodes and edges and move in accordance with their displacements.||The skeleton is deformed thanks to virtual forces that attract it towards the real position of the vessel represented by their center points.|