Epileptic seizure focus detection from interictal electroencephalogram: a survey

Abstract

Electroencephalogram (EEG) is one of most effective clinical diagnosis modalities for the localization of epileptic focus. Most current AI solutions use this modality to analyze the EEG signals in an automated manner to identify the epileptic seizure focus. To develop AI system for identifying the epileptic focus, there are many recently-published AI solutions based on biomarkers or statistic features that utilize interictal EEGs. In this review, we survey these solutions and find that they can be divided into three main categories: (i) those that use of biomarkers in EEG signals, including high-frequency oscillation, phase-amplitude coupling, and interictal epileptiform discharges, (ii) others that utilize feature-extraction methods, and (iii) solutions based upon neural networks (an end-to-end approach). We provide a detailed description of seizure focus with clinical diagnosis methods, a summary of the public datasets that seek to reduce the research gap in epilepsy, recent novel performance evaluation criteria used to evaluate the AI systems, and guidelines on when and how to use them. This review also suggests a number of future research challenges that must be overcome in order to design more efficient computer-aided solutions to epilepsy focus detection.

Keywords: Epilepsy; High-frequency oscillation (HFOs); Interictal electroencephalogram (EEG); Interictal epileptiform discharges (IEDs); Neural network; Phase amplitude coupling (PAC); Ripple and fast ripple; Seizure focus.

Fig. 1

An example of SOZ and EGZ of a surgical resection with seizure-free or seizure-persistent surgical outcomes

Fig. 2

Samples of focal and non-focal iEEGs (Bern Barcelona Dataset)

Fig. 3

Example of filtered ECoG signal in the low frequency range of (0.5–60 Hz) and in the ripple band (80–250 Hz). A16–A23 are the channel names. a is the sub-figure of the filtered ECoG signal for 10s in a low frequency range of (0.5–60 Hz). b is the filtered ECoG signal for 1s in the ripple band (80–250 Hz)

Fig. 4

Example of phase-amplitude comodulogram

Fig. 5

An example of IEDs (blue circle) in the SOZ channels (red color) as labeled by clinical experts. The iEEG data with a sampling rate of 2 kHz was measured in Juntendo University Hospital, Tokyo, Japan. (Color figure online)

Fig. 6

An unrolled recurrent network. $x_t$ and $y_t$ are the input and output at time t for each neuron. The input of each neuron contains not only the current input but also the output of the previous neuron

Fig. 7

A common neural network models. The top panel shows a traditional machine-learning model, which includes feature extraction and classification. The bottom panel illustrates the end-to-end machine-learning model that classifies directly from the input data

Fig. 8

The colormap representation on the left-hand side illustrates the channels (y-axis) and segment index (x-axis). Each yellow spot in the color map indicates detected focal segments. The right side of the color map (bar) represents the number of detected focal segments (x-axis) in each electrode (y-axis). The red bars indicate the SOZ channels labeled by epileptologists