Electrical impedance guides electrode array in cochlear implantation using machine learning and robotic feeder.

Area of Science:

• Biomedical Engineering
• Medical Devices
• Surgical Technology

Background:

• Cochlear implants restore hearing for profoundly deaf patients.
• Accurate electrode array (EA) placement in the scala tympani (ST) is crucial for optimal hearing outcomes.
• Current intraoperative methods for verifying EA placement are limited.

Purpose of the Study:

• To investigate the relationship between electrode array complex impedance and insertion trajectories.
• To develop a machine learning model for predicting EA placement during cochlear implant surgery.

Main Methods:

• A prototype system measured bipolar complex impedance (magnitude, phase, resistance, reactance) of electrodes.
• 137 insertions were performed into a plastic ST model using a 3-DoF actuation system at a speed of 0.08 mm/s.
• Machine learning algorithms (SVM, SNN) were used to classify full and partial insertion lengths and predict insertion direction.

Main Results:

• Support Vector Machine (SVM) achieved 97.1% accuracy for full insertion prediction.
• Shallow Neural Network (SNN) demonstrated 86.1% accuracy using partial insertion data (4 time samples, 2 apical electrode pairs) for real-time prediction.
• Direction prediction using partial data showed potential for online control of the insertion feeder.

Conclusions:

• Electrode array complex impedance can be a valuable indicator of insertion trajectory during cochlear implant surgery.
• Machine learning models, particularly SNN for real-time analysis, can accurately predict EA position.
• Online monitoring of EA placement via impedance holds promise for optimizing intraoperative positioning and improving patient hearing outcomes.