Fast parametric curve matching (FPCM) for automatic spike detection

Area of Science:

• Neurology
• Biomedical Engineering
• Signal Processing

Background:

• Epilepsy treatment often requires surgical resection of pathological cortical tissue.
• Accurate localization of epileptogenic zones is crucial for successful epilepsy surgery.
• Manual interictal spike detection in EEG/MEG data is time-consuming and may miss critical regions.

Purpose of the Study:

• To develop a novel, robust, and efficient automatic method for interictal spike detection.
• To improve the localization accuracy of epileptogenic cortical tissue for surgical planning.
• To address limitations of existing spike detection techniques, especially in low signal-to-noise ratio (SNR) conditions.

Main Methods:

• Proposed a biomimetic approach called Fast Parametric Curve Matching (FPCM).
• Constructed a constrained parametric morphological model based on peak-wave shape parametrization.
• Convolved the model with multichannel EEG/MEG data to determine spline parameters and detect spikes using logical predicates describing event morphology.

Main Results:

• FPCM demonstrated robustness and high AUC values in simulations under low SNR conditions compared to wavelet decomposition, template matching, and amplitude thresholding.
• Applied to human EEG/MEG and rat ECoG data, FPCM reliably detected interictal events.
• Localization of epileptogenic zones using FPCM was concordant with epileptologist's independent conclusions.

Conclusions:

• FPCM is a computationally efficient and flexible method for automatic interictal spike detection.
• The technique is tolerant to high-amplitude artifacts and adaptable to various target morphologies.
• FPCM offers a valuable tool to enhance the analysis of noisy interictal datasets for epilepsy surgery planning.