The event will take place on Monday, December 9, 2024 at 14h in the Hygie conference room, IHU Strasbourg.
Abstract
Cardiovascular diseases are one of the leading causes of death worldwide, with an incidence of 17.9 million deaths per year. The primary therapeutic solution is endovascular interventions, which owe their success to significant advantages, such as minimal invasiveness and low costs. However, these procedures are limited by their complexity, requiring extensive clinician training and access to specialized facilities. To deliver treatment, the clinician must navigate long, thin-tube devices, such as catheters and guidewires, through the patient's arteries, controlling them from the proximal end positioned at the entrance of the patient's vessel. This already challenging task is further complicated by the limited guidance the clinician has available, which is only provided by 2D fluoroscopic images. Furthermore, the acquisition of these images requires the use of X-rays, dangerous for the health of both the patient and the clinician, and the visibility of vessels is dependent on contrast agents, which can be harmful to the patient’s kidneys.
To address these limitations, this manuscript proposes solutions to facilitate the intervention, through the development of methods able to offer the clinician more support during certain phases of the intervention while automating others. A simplified procedure can indeed result in lower surgical time and, as a consequence, lower X-ray exposure for both the patient and the caregiver. Two main systems have been developed: one that enhances fluoroscopic images and another that autonomously navigates surgical tools. The first one is essentially an assistance system, which overlaps the classical fluoroscopic images with information about the anatomy that is being navigated and shows the clinicians the predicted outcome of their actions before they perform them. The second one is a Deep Reinforcement Learning controller which aims to autonomously perform the procedure by controlling an endovascular surgical robot. Currently, these robots function only as leader-follower devices, which are not able to provide additional support to the caregiver during the procedure.
In tests conducted on a phantom in the context of a user study, the enhanced fluoroscopic system allowed for a 56% intervention time reduction. The autonomous controller achieved a success rate of over 95%, even when tested on anatomies with characteristics completely different from the training models.
Jury Composition:
Reviewers:
- Pr. Franziska Mathis-Ullrich, University of Erlangen–Nuremberg
- Pr. Emmanuel Vander Poorten, University of KU Leuven
Examiners :
- Pr. Philippe Cattin, University of Basel
Supervisors:
- - Pr. Stéphane Cotin, Inria Strasbourg, University of Strasbourg
- - Pr. Florent Nageotte, University of Strasbourg
- - Pr. Michel Duprez, Inria Strasbourg, University of Strasbourg
