Fast and efficient rejection of background waveforms in interictal EEG

Abstract

Automated annotation of electroencephalograms (EEG) of epileptic patients is important in diagnosis and management of epilepsy. Epilepsy is often associated with the presence of epileptiform transients (ET) in the EEG. To develop an efficient ET detector, a vast amount of data is required to train and evaluate the performance of the detector. Interictal EEG data contains mostly background waveforms, since ETs only occur occasionally in most patients. In order to detect ETs in an automated fashion, it is meaningful to first try to eliminate most background waveforms by means of simple, fast classifiers. The remaining waveforms can in a following step be processed by more sophisticated and computationally demanding classification algorithms, such as deep learning systems. In this study, we design a cascade of simple thresholding steps to reject most background waveforms in interictal EEG, while maintaining most ETs. Several simple and quick-to-compute EEG features are chosen. By thresholding these features in consecutive steps, background waveforms are rejected sequentially. In our numerical experiments, a cascade of 10 steps is able to reject 98.65% of all background segments in the dataset, while preserving 90.6% of the ETs.

Interictal epileptiform discharge characteristics underlying expert interrater agreement

“…Some previous studies have applied nonlinear energy operator (NLEO) for IED detection purposes (Mukhopadhyay and Ray, 1998; Liu et al, 2013; Bagheri et al, 2016). The NLEO is computed as: $${\phi }_k\{\mathrm{x}[\mathrm{n}]\}={\mathrm{normalx}}^2[\mathrm{n}]-\mathrm{x}[\mathrm{n}-\mathrm{k}]\mathrm{x}[\mathrm{n}+\mathrm{k}],$$where the resolution parameter k (Choi and Kim, 2002) is chosen to have values in the range of 1–40.…”

“…Many recent IED detection algorithms have applied wavelet analysis (Senhadji et al, 1995; Latka et al, 2003; Güler and Übeyli, 2005; Indiradevi et al, 2008; Bagheri et al, 2016). In this study, the discrete wavelet transform (DWT) was applied to decompose the signal into multiple scales based on wavelet basis functions.…”

Classifier Cascade to Aid in Detection of Epileptiform Transients in Interictal EEG

“…Some methods have combined different techniques of classification as well as artifact removal methods [10–13]. For instance, they have applied template matching combined with clustering [10], template matching combined with support vector machines (SVMs) and independent component analysis (ICA) [11], nonlinear energy operator combined with mimetic analysis and Adaboost classifiers [12], and sequence merging combined by SVMs [13], wavelet transform with machine learning techniques [14, 15]. In a recent study [16], the performance of an ET detection software on a dataset consisting of 40 epileptic patients, was compared with ET detection by three skilled technologists.…”

Epilepsy diagnosis from scalp EEG : a machine learning approach

“…The background rejection system separated approximately 1% of the background waveforms, with less than 1% loss in sensitivity. The performance of the background rejection system is superior to the different methods suggested in the literature [131]. Later the extracted FP backgrounds were applied for improving the performance of the various IED detectors.…”

“…After the rejection, sophisticated and highly accurate techniques are applied to the remaining miniature dataset. Bagheri et al has demonstrated the method of background rejection in [131]. The feature set consisted of morphological features, Non-Linear Energy Operator (NLEO) values, Discrete Wavelet Transform (DWT) coefficients, Continuous Wavelet Transform (CWT) coefficients, and frequency band features.…”