Multicontact Localization Framework for Flexible Robots Using a Data-Driven Approach

Abstract

Minimally invasive surgery (MIS) is increasingly employed in several medical disciplines. During minimal invasive cardiovascular approaches, flexible instruments, such as catheters or endoscopes, are navigated along a tortuous path to reach a deeply seated target site. Navigation through the vasculature is not without risk, as the vessel can be fragile, potentially with plaque or calcification that could get dislodged if excessive stresses are caused during navigation. Knowledge of the interaction forces between these flexible tools and the vasculature could help steer the instruments and avoid exerting excessive forces on the surrounding tissues. Integrating force sensing into the tip of the instruments has been proposed in the art. However, tip force sensing is blind to forces that arise along the catheter body’s length. This article proposes a data-driven framework to estimate the locations of multiple contact forces along the catheter’s length. The proposed approach consists of a contact state detection and, subsequently, a contact localization method. Both methods make use of knowledge of the shape of the catheter’s body. This shape is measured here by a multicore fiber Bragg grating (FBG) fiber. The proposed method also works for steerable catheters, as it allows accounting for the effect of the actuation of the catheter itself. A state-of-the-art contact localization method using multicore FBG fiber is implemented and serves as a baseline to compare the newly proposed method with. The proposed data-driven approach requires a minimum amount of training data, which can be collected by simply actuating the robotic catheter in free space over its workspace. The method is validated with a commercial MitraClip steerable guide catheter. Dynamic experiments show that the proposed contact state detection method can detect contact in 0.44 s. The average contact localization errors of the proposed method and state-of-the-art method are 6.1 mm (1.6%) and 18.0 mm (4.9%), respectively.