FFT Power Relationships Applied to EEG Signal Analysis: A Meeting between Visual Analysis of EEG and Its Quantification

Read the full article See related articles

Listed in

This article is not in any list yet, why not save it to one of your lists.
Log in to save this article

Abstract

Objective

This study presents a novel computational approach for analyzing electroencephalogram (EEG) signals, focusing on the distribution and variability of energy in different frequency bands. The proposed method, FFT Weed Plot, systematically encodes EEG spectral information into structured metrics that facilitate quantitative analysis.

Methods

The methodology employs Fast Fourier Transform (FFT) to compute the Power Spectral Density (PSD) of EEG signals. A novel encoding technique transforms frequency band distributions into six-entry vectors, referred to as “words,” which serve as the basis for three key metrics: a scalar value a vector , and a matrix H . These metrics are evaluated using a dataset comprising EEG recordings from 30 healthy individuals and 15 patients with epilepsy. Machine learning classifiers are then applied to assess the discriminatory power of the proposed features.

Results

The classification models achieved a 95.55% accuracy, 93.33% sensitivity, and 96.67% specificity, demonstrating the robustness of the proposed metrics in distinguishing between control and epileptic EEGs.

Conclusions

The FFT Weed Plot method provides a novel approach for EEG signal quantification, improving the systematization of spectral analysis in neurophysiological studies. The metrics developed could serve as quantitative descriptors for automated EEG interpretation, offering potential applications in clinical and research settings.

Highlights

  • From frequency domain analysis to information and probability theory, new ways of encoding information.

  • A step towards the systematization and automation of medical EEG reading.

  • New global metrics for the description of the energy of an EEG recording and their applications in machine learning.

  • The FFT Weed Plot method, We present a new, reproducible, robust and clinically designed method to improve the objectivity of medical practice and research in neurophysiology.

Article activity feed