Intelligent image guidance in cardiac interventions

pIMAGIC1In complex minimally invasive interventions, image guidance and visualization are often bottlenecks in the feedback loop of physician, instrument and patient. This project investigates generalized intelligent instrument & anatomy tracking and visualization to eliminate feedback limitations. These innovative concepts will be applied in two complex and clinically highly relevant minimally invasive procedures: radiofrequency ablation (RFA) treatment of atrial fibrillation, imaged with ultrasound, and treatment of chronic total occlusions (CTO) in coronary arteries, imaged with X-ray. In these procedures, novel fiber optics, ultrasound and instrument steering come together, in close collaboration with the corresponding iMIT projects


A generalized image guidance platform for steerable interventional instruments will be developed that will introduce 3D models of anatomy and anatomic motion, as well as instrument models into the feedback chain. Static alignment of 3D models to 2D (projection) images is state of the art. However, in cardiac interventions this alignment is invalidated by respiratory and cardiac motion and deformation. We will merge (novel) real-time images, tracked instruments, and pre-operative 3D and 4D data during cardiac interventions, to improve the visualization.

WP1.2.1. RFA treatment of atrial fibrillation, imaged with intracardiac US (Erasmus MC)

We will use a 2D intracardiac echocardiography (ICE) catheter with 3D position sensing to image the ablation region and surrounding anatomy. By mechanically manipulating the ICE catheter, fast 3D sweeps of the ablation zone can be created [3]. By fitting the images to a 3D multi-cavity anatomical model, incremental 3D image reconstruction, fusion with preoperative CT/MR images and electro-anatomical maps is possible. Live instrument tracking will be realized by an image-based Bayesian tracking approach. Integration of photoacoustic imaging of the ablation (Project 2.1) within the 3D reconstruction will supply direct feedback of the efficacy of the ablation.

WP 1.2.2. Treatment of chronic total occlusions (CTO) in coronary arteries (Erasmus MC)

We will integrate 3D anatomical information from CTA in fluoroscopic images in order to visualize the occluded segment from CTA, including additional relevant information such as calcifications. An initial alignment can be obtained by registering CTA information to interventional angiograms. We will develop and evaluate novel techniques that maintain the alignment after the initial registration. We will develop image-based tracking of structures, focusing on instruments such as catheters and guide wires as landmarks. Subsequently, learning-based approaches, utilizing the image information will be developed to track the location of the target lesion, enabling continuous and accurate visualization of the target anatomy. We will focus both on the development of fast and robust extraction of landmarks, and the development and evaluation of machine learning approaches that allow learning the relation between these landmarks and the target structures. In both stages, we will utilize 3D and 4D anatomical models obtained from pre-operative data. For the development and evaluation of the algorithms, a database with pre-operative CTA and interventional images will be built.