Skip to Main content Skip to Navigation
Theses

Image-guided Interactive Simulation for Endovascular Surgery

Raffaella Trivisonne 1, 2, 3
3 MAGRIT - Visual Augmentation of Complex Environments
Inria Nancy - Grand Est, LORIA - ALGO - Department of Algorithms, Computation, Image and Geometry
Abstract : Minimally invasive fluoroscopy-based procedures are the gold standard for diagnosing and treating various pathologies of the cardiovascular system. With this kind of procedures, clinicians have to infer the 3D shape of the device from 2D images. Such lack of depth perception, combined with a dense environment of overlaying anatomical structures, has been identified as one of the major factors affecting clinical performances. Several methods have been proposed to enhance the visualization of 2D fluoroscopic images, which could improve the clinician’s global insight and consequently the positive outcomes of the procedures. A widely used approach is to create a 3D reconstruction of the surgical scene to be combined with 2D fluoroscopic images, in order to have an augmented view. In general, this kind of methods aims at retrieving the 3D shape of the device (and or anatomical structures) by combining some priors on the shape and the behaviour of the device, with external observations, providing some incomplete information on its current state. After highlighting the limitations of the existing 2D-3D reconstruction methods, our objective was to develop a method that: 1. provides a good description of both the shape and the behavior of the device; taking into account non-rigid interactions with the surrounding anatomy and non-linear phenomena (e.g. non-sliding contacts); 2. solely relies on monocular 2D fluoroscopic images, without the need to embed any external sensors onto the interventional device; 3. takes into account and compensates the uncertainties which might exist on model parameterization and the noise affecting external observations; 4. is compatible with real-time computations; We first proposed a purely deterministic approach, where projective information from 2D fluoroscopic images is integrated to the model as mechanical constraints. Despite the good results, the proposed method is not able to take into account non-linear phenomena such as stick and slip transitions. In addition, errors on both the navigation model and external observations are not taken into account. For the above reasons, we designed a new stochastic approach. Given the ill-posedness of the 2D-3D reconstruction problem, the 3D shape of the interventional device can be seen as a random variable. Such variable is described, at the same time, through an process model, which provides a description of the variable through time and it is affected by some uncertainties, and some external observations, which can provide some partial information on its current configuration and are affected by noise. In particular, this thesis aims to develop a novel approach, where a Bayesian approach is used to combine a constrained physics-based simulation of the catheter navigation, with external 2D observations extracted from 2D fluoroscopic images. Whereas the physics-based model provides a prediction of the shape of the navigation device navigating the blood vessels (taking into account non-linear interactions between the catheter and the surrounding anatomy), an Unscented Kalman Filter is used to correct the navigation model using 2D features, extracted from fluoroscopic images, as external observations. The proposed method has been evaluated on both synthetic and real data. Lastly, we present and analyse the current limitations of our method, proposing possible solutions, along with some perspectives for future works and applications.
Complete list of metadatas

https://hal.inria.fr/tel-03021043
Contributor : Raffaella Trivisonne <>
Submitted on : Tuesday, November 24, 2020 - 10:34:13 AM
Last modification on : Friday, December 4, 2020 - 3:45:40 PM

File

Trivisonne_Raffaella_2020_ED26...
Files produced by the author(s)

Identifiers

  • HAL Id : tel-03021043, version 1

Citation

Raffaella Trivisonne. Image-guided Interactive Simulation for Endovascular Surgery. Modeling and Simulation. Université de Strasbourg, 2020. English. ⟨tel-03021043⟩

Share

Metrics

Record views

111

Files downloads

65