Method for predicting epileptic seizure earlier stages on basis of EDM (empirical mode decomposition) and WPD (wavelet packet decomposition) feature fusion

Abstract

The invention discloses a method for predicting epileptic seizure earlier stages on the basis of EDM (empirical mode decomposition) and WPD (wavelet packet decomposition) feature fusion. The method includes steps of 1, dividing categories of EEG (electroencephalography) signals, cutting samples, carrying out empirical mode decomposition on each sample, analyzing first intrinsic mode functions obtained by means of decomposition, and extracting features on the basis of the FIMF (first intrinsic mode functions) from the first intrinsic mode functions; 2, carrying out four-layer wavelet packet decomposition on the first intrinsic mode functions obtained at the step 1 to obtain wavelet packet coefficients of 16 nodes of every last layer and extracting the skew, energy and features on the basisof FIMF-WPD from the wavelet packet coefficients; 3, fusing the features on the basis of the FIMF and the features on the basis of the FIMF-WPD and building epileptic seizure earlier stage predictionmodels for fused feature training classifiers by the aid of random forest algorithms. The features on the basis of the FIMF comprise energy, energy ratios and variance. The method has the advantages that time periods of the epileptic seizure earlier stages can be effectively warned by the aid of the method, and accordingly secondary injury on patients due to epileptic seizure can be reduced.

Description

technical field [0001] The invention belongs to the field of digital signal processing and smart medical treatment, and relates to an accurate detection method for epileptic seizures and pre-seizures based on Empirical Mode Decomposition (EMD) and Wavelet Packet Decomposition (WPD). Background technique [0002] The traditional epilepsy detection method based on EEG signal processing is usually used to solve the binary classification problem, that is, the signal is divided into seizure period and non-ictal period; or the three-category problem, that is, the signal is divided into seizure period, preictal period and seizure period. intermission. Compared with the traditional detection of epileptic seizures, accurate prediction of pre-seizures is more important to effectively reduce the secondary harm that may be caused by epileptic seizures. However, there is currently no algorithm specifically for pre-seizure detection. In order to solve this problem, the present invention ...

AI Technical Summary

Problems solved by technology

Compared with the traditional detection of epileptic seizures, the accurate prediction of pre-seizures is more important to effectively re

Images