Towards Exoscope Automation in Neurosurgery: A Markerless Visual-Servoing Approach

Abstract

Exoscopes are a promising tool for neurosurgeons, offering improved visualisation and ergonomics compared with traditional surgical microscopes. They consist of an external scope that projects the surgical field onto a 2D or 3D monitor, providing a wider field of view and better access to the surgical site. Despite the advantages, exoscopes present some limitations, such as the need for manual or foot joystick repositioning, which can disrupt the flow of the procedure and increase the risk of user error. In this study, a markerless visual-servoing approach for autonomous exoscope control is proposed to address these limitations and enhance the ergonomics and reduce the physical and cognitive load compared with traditional joystick control. The system uses visual information from the operating field to control the exoscope, eliminating the need for markers or additional tracking devices. The proposed approach was validated using a 7-DOF robotic manipulator with a stereo camera in an eye-in-hand configuration. Results showed that the system achieved 89 % accuracy in detecting the target and tracking its movement with a tracking error ranging from 0.50±0.17 cm for low-speed movements to 1.38±0.73 cm for high-speed movements. The proposed system also demonstrated improved efficiency, with a shorter execution time of 72.07±19.36 s compared with 106.52±18.50 s for the foot-joystick control. Additionally, the time out of the FoV was significantly higher in the joystick control mode and the frequency of appearance of the instrument in the centre of the image was higher when using the proposed system. The NASA TLX results indicated lower physical and cognitive load compared with the joystick control-based modality.