Robot-assisted perception augmentation for online detection of insertion failure during cochlear implant surgery

Abstract

SUMMARY During the past decade, robotics for cochlear implant electrode array insertion has been limited to manipulation assistance. Going beyond manipulation assistance, this paper presents the new concept of perception augmentation to detect and warn against the onset of intracochlear electrode array tip folding. This online failure detection method uses a combination of intraoperative electrode insertion force data and a predictive model of insertion force profile progression as a function of insertion depth. The predictive model uses statistical characterization of insertion force profiles during normal robotic electrode array insertions as well as the history of intra-operative insertion forces. Online detection of onset of tip folding is achieved using the predictive model as an input into a support vector machine classifier. Results show that the detection of tip folding onset can be achieved with an accuracy of 88% despite the use of intra-operative insertion force data representing incomplete insertion. This result is significant because it allows the surgeon or robot to choose a corrective action for preventing intra-cochlear complications.