Endovascular Surgery Fusion

Project Info

Fusion Project

Fundings IHU Project in collaboration with Siemens

Objective Elastic registration of a virtual model of the targeted organ on fluoroscopic X-ray images

Date 02. 2017 — 02. 2019

Contact antoine.a.petit@inria.fr,  bruno.josue.marques@inria.fr


Project Description

Endovascular surgery is a medical specialty of minimally invasive procedures which treats pathologies affecting the blood vessels relying on the use of catheters, guide-wires and other endovascular devices.
During the intervention, the clinician relies entirely on fluoroscopic images, obtained through a X-ray beam, allowing to display in real time the 2D projection of the anatomy of the patient, the current position of the catheter and its displacements within the blood vessels.
Despite the evident benefits of this technique, the exposure of both clinicians and patients to cumulative doses of ionizing radiation is an important issue. A contrast medium – which is usually allergenic and nephrotoxic – needs to be injected regularly in the vascular system in order to visualize blood vessels through which the instruments are inserted.
In addition, fluoroscopic images are characterized by a lack of depth perception and a poor quality of visualization, due to overlaying structures within the grey-scale image.
Using Augmented Reality to enhance this scenario, by combining the 2D fluoroscopic images with a 3D model of the vessel structures in which the radiologist operates would enhance the perception of depth, which would have many benefits, both for the patient and the surgeon:
– Reducing the operating time, by simplifying navigation
– Locating organ’s anomalies during the procedure
– Reducing the amount of radiation absorption both for the patient and for the radiologist
– Reducing the amount of contrast medium
The project’s main challenge is to register in real time an elastic, patient specific model of the targeted organ onto the intraoperative images. To broaden the clinical use of this project, we want to address this problematic by using deep learning approaches, computer vision and physics-based simulation.

Project Gallery


Coming up!

People Involved