FBG-based Actuation and Data Driven Contact Detection for Smart Steerable Instruments

Abstract

Catheters and guidewires are increasingly used to navigate tortuous paths offering minimal invasive access to deeply seated locations in the body. Steering these instruments is highly challenging among others due to poor awareness of the configuration such instrument takes on in the body. To address this difficulty research in physical intelligence has been conducted. The aim is to delegate part of the control problem locally and have the instrument determine itself how to effectively interact with its physical environment. To enable such smart behaviour this paper presents a compact FBG (fiber Bragg grating) based drive system for controlling the bending of the distal tip of a steerable catheter. The design process establishes key constraints for selecting an appropriate FBG fiber based on the selected backbone’s characteristics. Force estimation is done using the strain measured via the fiber with a root mean square error (RMSE) of 0.05 N which is then used to train a Long Short-Term Memory (LSTM) network to detect possible contact with the surroundings using the force prediction of the trained model. The trained model was able to predict the force with an RMSE of 0.012 N in a non-contact scenario. The results indicate that the proposed system incorporating FBG sensing, pneumatic artificial muscle (PAM) actuation, and LSTM based contact detection offers a promising pathway for more precise and versatile catheter manipulation in minimally invasive interventions.