Comparison of Performance Metrics for Real-Time Haptic Feedback in Surgical Skill Training

Abstract

Endovascular surgery involves minimally invasive sur- gical techniques that can result in significantly shorter operation times and hospital stays, lower complication rates, less blood loss, and lower rates of postoperative mechanical ventilation and atrial fibrillation than the equivalent open procedures [1], [2]. Repeated practice is central to skill acquisition, and minimally invasive procedures like endovascular surgery may require more or specialized practice compared to traditional surgery. For example, despite known benefits of endovascular aortic valve replacement compared to traditional surgical methods, Smith et al. attributed observations of higher rates of stroke, transient ischemic attacks, and major vascular complications to a protracted learning curve for the endovascular approach [3]. Virtual reality endovascular surgical simulators can be loaded with a patient’s pre-operative CT scan, enabling rehearsal of difficult cases before operating. Simulators are also accessible to trainees, giving opportunities for additional practice in navigating to hard-to-reach vas- cular structures, or exposure to rare procedures. Still, surgical simulators lack the provision of real-time and objective performance feedback. Instead, feedback is only available after the completion of a surgical task, and often does not provide the trainee with insight into how they should change their task performance strategies to achieve performance goals. Objective measures of skill derived from endovascular guidewire movement kinematics that characterize tool tip movement smoothness have been shown to correlate with expertise [4], [5]. Such metrics have not yet been used during training as real-time performance feedback, despite evidence that providing feedback can improve training outcomes [6]. Our approach to providing real-time performance feed- back during surgical skill training is intended to address this gap. We propose to use estimates of spectral arc length (SPARC), idle time, and average velocity to quantify task performance, then encode these measures as vibrotactile cues displayed to trainees in a wearable haptic device (see Fig. 1).