Estimation of Contact Forces of Endovascular Devices Using Physicians' Bio-Signals

Cardiovascular diseases are one of the most common death causes. Endovascular interventions, such as the placement of stents via a catheter-based procedure, can often be life-saving. Such interventions are technically challenging, and physicians require several years of training before achieving competence and certification. Due to the high number of patients and the lack of properly trained staff, robotic solutions for, e.g., stent placement could improve access to care, facilitating the treatment of more patients in both high- and low-income countries. However, such solutions must also respect the endovascular devices' operational ratings, which often limit the safe contact forces between the device and surrounding tissues within a threshold established by experienced physicians. Yet, the exact definition of these excessive force limits remains unclear. To inform the development of safety mechanisms for robotics solutions, it is essential to first understand the typical force interactions during manual procedures. Given the impracticality of embedding sensors across the diverse range of endovascular tools, we propose a data-driven approach to estimate these forces. In this work, we present a novel method for quantifying contact forces during interventions by analyzing the physician's muscular activity and motion patterns. The setup is characterized by a series of tests, in which we demonstrate the feasibility of tracking contact forces during the insertion of an endovascular instrument, into a phantom model. Our results reveal that we can estimate forces within the Weber's fraction for the upper arm, relative mean absolute error < 13%, implying that the force discrepancies between the actual and estimated values are unnoticeable to a human user.Clinical relevance-As an alternative to the direct measurement of interaction forces, this work proposes a novel way to measure forces exerted on an endovascular device inserted in a phantom using only wearable sensors. The main advantage of this approach is that it enables the use of standard, non-sensorized equipment, essentially converting the operator themselves into a sensor system.