Robust compression and detection of epileptiform patterns in ECoG using a real-time spiking neural network hardware...

Area of Science:

• Neuroscience
• Neuromorphic Engineering
• Computational Biology

Background:

• Interictal Epileptiform Discharges (IED) and High Frequency Oscillations (HFO) detected via intraoperative electrocorticography (ECoG) are crucial for identifying the epileptogenic zone during epilepsy surgery.
• Real-time processing of ECoG signals is essential for immediate surgical guidance.

Purpose of the Study:

• To develop and validate a real-time system for detecting HFO and IED-HFO using a spiking neural network (SNN) on a neuromorphic device.
• To assess the system's performance against established offline algorithms and its potential for clinical application.

Main Methods:

• A modular spiking neural network (SNN) was designed and implemented on a mixed-signal neuromorphic device.
• The SNN was interfaced with the BCI2000 real-time framework for processing intraoperative ECoG data.
• The system was validated using pre-recorded data and a remote online analysis of ECoG signals.

Main Results:

• The SNN achieved HFO detection rates concordant with a validated offline algorithm (Spearman's ρ = 0.75, p = 1e-4).
• The system demonstrated identical postsurgical seizure freedom predictions compared to the offline method for all patients.
• Successful real-time IED-HFO detection was achieved in a remote online analysis, demonstrating data compression and transfer capabilities.

Conclusions:

• The developed SNN system provides accurate and real-time detection of HFO and IED-HFO from ECoG signals.
• This technology shows significant promise for enhancing intraoperative decision-making in epilepsy surgery.
• Automated remote real-time detection of these biomarkers can facilitate their broader clinical adoption.