Feature extraction and recognition of epileptiform activity in EEG by combining PCA with ApEn

Abstract

This paper proposes a new method for feature extraction and recognition of epileptiform activity in EEG signals. The method improves feature extraction speed of epileptiform activity without reducing recognition rate. Firstly, Principal component analysis (PCA) is applied to the original EEG for dimension reduction and to the decorrelation of epileptic EEG and normal EEG. Then discrete wavelet transform (DWT) combined with approximate entropy (ApEn) is performed on epileptic EEG and normal EEG, respectively. At last, Neyman-Pearson criteria are applied to classify epileptic EEG and normal ones. The main procedure is that the principle component of EEG after PCA is decomposed into several sub-band signals using DWT, and ApEn algorithm is applied to the sub-band signals at different wavelet scales. Distinct difference is found between the ApEn values of epileptic and normal EEG. The method allows recognition of epileptiform activities and discriminates them from the normal EEG. The algorithm performs well at epileptiform activity recognition in the clinic EEG data and offers a flexible tool that is intended to be generalized to the simultaneous recognition of many waveforms in EEG.

Flexible, Scalable and Energy Efficient Bio-Signals Processing on the PULP Platform: A Case Study on Seizure Detection

“…The best performance is obtained in the LDA feature extraction case. The work in [8] shows a method to recognize epileptic activity through a feature extraction approach based on PCA to reduce the dimension of the data, DWT to extract frequencies at a certain resolution level and ApEn to estimate the quantity of entropy contained in the signal (a low value indicates determinism, while a high value expresses randomness). After this, a threshold is established to be applied to the ApEn values.…”

“…In contrast with the previously-presented solutions, exploiting an optimized near-threshold micro-architecture and the most advanced FD-SOI technology, the PULP platform allows achieving high performance and energy efficiency, coupled with the high versatility of programmable processors. Although the proposed implementation has been applied to a dataset with 23 electrodes due to the availability of the dataset, in the context of seizure detection applications, the availability of a fully-programmable platform allows trading the detection latency with a larger number of electrodes, addressing the key challenges highlighted by the trend in research, which goes toward the design of dense multichannel systems employing up to 128 electrodes [6][7][8]. Furthermore, system programmability is preferable to deal with the processing chains typical of most other biomedical applications, which need to be often updated or tuned during the life-time of a system [26].…”

“…((float)y/(float)(1 << FWL)) (8) where y is a fixed point number. Sums and subtractions are performed as a common integer operation; the result of a sum or a subtraction between two Q13.19 numbers is a Q13.19 number, while multiplications and divisions are more challenging.…”

“…While the trend in research goes toward the design of dense multichannel systems [6][7][8] with large and dense arrays of sensors (i.e., up to 128 electrodes) to allow a fine-grain coverage of the brain surface and target wider areas, current wearable and implantable solutions for seizure detection are only able to manage a few electrodes with a latency of 500 ms due to their limited computing power [9,10]. The most common deeply-embedded systems for seizure detection are based on machine learning that can achieve high accuracy with a relatively small computational effort, for a limited number of electrodes.…”

PCA Based EEG Event Features Extraction for Robotics Dexterous Grasping

“…al. [8] with some modifications. For a matrix (X) made of EEG signals (xk), where k=1, 2, 3, …, n, assuming (n) rows where each of the signals contains (m) samples as given by Eq (1).…”