Real-Time Insertion Depth Tracking of Cochlear Implant Electrode Array With Bipolar Complex Impedance and Machine Intelligence

Abstract

Cochlear implants have significantly improved hearing for many as the most successful prosthesis, however, hearing outcomes vary. Uncertainty during electrode array (EA) insertion, including trauma and depth control, is one factor. To minimize radiation exposure from imaging methods like CT scans, this in-vitro study investigates the use of bipolar electrode impedance and artificial intelligent models to determine EA insertion depth. Complex impedance data was collected by inserting a commercial EA into a scaled-up 2D scala tympani model using a robotic feeder system. A support vector machine model produced a 98% classification accuracy for final insertion depth estimation. A CNN-LSTM hybrid model yielded 0.85 R-squared and 1.72 mm mean absolute error in depth estimation at each millimeter during a 25 mm insertion. This approach to depth assessment based on impedance may help with cochlear implant procedures and find use in other medical implant applications.