The clinical investigation evaluates whether speech can be decoded from neural activity recorded with cortical surface electrocorticography (ECoG) electrodes during awake brain tumor surgery. Neural signals and voice recordings are collected while patients perform a picture naming task as part of standard intraoperative mapping. The study assesses the ability of machine learning algorithms to predict the named item from intraoperative electrophysiological recordings.
Patients with severe speech and motor impairments often lose the ability to communicate, detrimentally affecting their quality of life. Invasive brain-computer interfaces (BCIs) have shown promise in restoring lost speech functionalities by decoding neural activity of intended or attempted speech. However, the systems with the most promising results rely on highly invasive micro-electrode arrays that tend to irritate the brain tissue leading to inflammation, tissue scarring and ultimately, loss of signal. A promising long-term alternative is intracranial electrode strips/grids that rest on the surface of the brain without having to penetrate the cortex. The contact density of standard strips/grids used for electrophysiological mapping is very low (1 contact/cm2 compared to 100 contacts/cm2), however, a new generation of electrode strips/grids offer decent contact density closing the distance with micro-electrode arrays. In order to determine if the contact density of cortical surface electrode grids is sufficient for decoding speech, we aim to record and decode neural activity of tumor patients performing a picture naming task during awake surgeries, as these operations offer the only opportunity to test the use of cortical surface electrodes for neural speech decoding.
Study Type
INTERVENTIONAL
Allocation
RANDOMIZED
Purpose
OTHER
Masking
NONE
Enrollment
20
INBRAIN Graphene Cortical Interface (Non-CE-marked Class III )high-density graphene ECoG grid placed on cortical surface for up to 2 hours during awake surgery. Recording only, no stimulation.
INBRAIN Graphene Cortical Interface (Non-CE-marked Class III )high-density graphene ECoG grid placed on cortical surface for up to 2 hours during awake brain surgery. Recording only, no stimulation.
Department of neurosurgery Lariboisière hospital-APHP
Paris, Île-de-France Region, France
Primary outcome Measure
The primary metric for evaluating the success of the project is the accuracy of the machine learning algorithms in decoding neural activity to predict the named item by the patient. Achieving an Area Under the Curve (AUC) value of greater than 0.9 indicates high predictive performance.
Time frame: Up to 12 months after the last patient's intraoperative recording
Secondary Outcome Measure
The evaluation criterion is to identify axono-cortical evoked potentials (ACEPs) with a signal- to-noise ratio \> 1. The parameters consist of all recordings made by surface electrodes during axonal electrical stimulation at 1 Hz. On each channel, averaging will be performed, synchronized with the 1 Hz stimulation.
Time frame: Up to 12 months after the last patient's intraoperative recording
This platform is for informational purposes only and does not constitute medical advice. Always consult a qualified healthcare professional.
DIXI C10-12BIOM and WISE WCS00-4S10-000 (CE-marked Class III) low-density ECoG strips placed on cortical surface for up to 2 hours during awake surgery. Recording only, no stimulation.