Classification of epileptic EEG recordings using signal transforms and convolutional neural networks

San-Segundo, Rubén; Gil-Martín, Manuel; D’Haro, Luis Fernando; Pardo, José Manuel

doi:10.1016/j.compbiomed.2019.04.031

Cited by 135 publications

(68 citation statements)

References 47 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…That provides a more comparable framework, given that the use of works using other datasets may expose significant differences when performing the same classification method. As an example, San-Segundo et al [25] shows that the focal-nonfocal (F-NF) classification accuracy may differ more than 20% when the same methods applies to the Bern-Barcelona dataset [7] and the Epileptic Seizure Recognition dataset [46]. In this example, the nature of the signals in the dataset, mainly the difference between signal lengths (only 1s for the latter), makes the difference.…”

Section: Discussionmentioning

confidence: 99%

“…However, these works tend to require a considerable computational load, especially the most recent ones. As an example, San-Segundo et al [25] proposed a deep neural network (DNN) made up of two convolutional layers for feature extraction and three fully connected layers for classification. In this work, authors increased the classification accuracy a little at the expense of increasing, considerably, its computational needs.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Real-Time Localization of Epileptogenic Foci EEG Signals: An FPGA-Based Implementation

et al. 2020

View full text Add to dashboard Cite

The epileptogenic focus is a brain area that may be surgically removed to control of epileptic seizures. Locating it is an essential and crucial step prior to the surgical treatment. However, given the difficulty of determining the localization of this brain region responsible of the initial seizure discharge, many works have proposed machine learning methods for the automatic classification of focal and non-focal electroencephalographic (EEG) signals. These works use automatic classification as an analysis tool for helping neurosurgeons to identify focal areas off-line, out of surgery, during the processing of the huge amount of information collected during several days of patient monitoring. In turn, this paper proposes an automatic classification procedure capable of assisting neurosurgeons online, during the resective epilepsy surgery, to refine the localization of the epileptogenic area to be resected, if they have doubts. This goal requires a real-time implementation with as low a computational cost as possible. For that reason, this work proposes both a feature set and a classifier model that minimizes the computational load while preserving the classification accuracy at 95.5%, a level similar to previous works. In addition, the classification procedure has been implemented on a FPGA device to determine its resource needs and throughput. Thus, it can be concluded that such a device can embed the whole classification process, from accepting raw signals to the delivery of the classification results in a cost-effective Xilinx Spartan-6 FPGA device. This real-time implementation begins providing results after a 5 s latency, and later, can deliver floating-point classification results at 3.5 Hz rate, using overlapped time-windows.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Real-Time Localization of Epileptogenic Foci EEG Signals: An FPGA-Based Implementation

et al. 2020

View full text Add to dashboard Cite

show abstract

“…Scholars all over the world have extensively studied EMD and applied it to various fields, such as medicine, machinery, and geography. Rubén applied EMD to process and analyze physiological characteristic signals, which are beneficial for the pathological study of diseases [6]. EMD was also used to decompose seismic waves to obtain the intrinsic information of signals [7], which provided technical support for effectively identifying earthquake levels.…”

Section: State Of the Artmentioning

confidence: 99%

“…Moreover, wavelet transform is most suitable for processing with low-frequency components [5]. By contrast, empirical mode decomposition (EMD) could display a full view of signals in the time and frequency domains simultaneously, especially during treatment of non-stationary and nonlinear signals [6][7][8][9][10]. Rough sets are a kind of mathematical tool to describe the incompleteness and uncertainty of data.…”

Section: Introductionmentioning

confidence: 99%

Fault Diagnosis Based on the Optimization of Characteristic Parameters and Neural Networks of Gearboxes

Liu¹,

Zhang²

2019

JESTR

View full text Add to dashboard Cite

Gearboxes are the most commonly used transmission components in heavy equipment such as helicopters, shearers, and ships. The failure rate of gearboxes is high, and the characteristic signals under faulty conditions tend to be extremely weak and are often overwhelmed by strong noise. Thus, extracting sensitive characteristic parameters is difficult. In order to optimize the characteristic parameters of gearboxes and improve diagnosis efficiency, this study proposed a method for fault diagnosis of gearboxes that combines empirical mode decomposition (EMD) with rough sets and neural networks. First, the principle of EMD was explored. The indicators for measuring characteristic parameters were identified to compare the feature set composed of energy values with those comprising approximate entropy parameters. Second, the conditional attribute reduction technique for rough sets was investigated. An algorithm for attribute reduction based on conditional equivalence classification was put forward for parameter optimization. Then, a neural network was employed to identify the feature sets before and after the attribute reduction. Results show that the energy characteristic set is the most sensitive to failures. The attribute reduction technique reduces the characteristic parameters from 6 to 4, thereby effectively lowering the input vectors of the neural network. The training time is also decreased from 1.024 s to 0.351 s, obviously promoting the efficiency of fault diagnosis. The study provides references for improving the performance of online real-time fault diagnosis.

show abstract

“…Psychosis develops 14.4 years after onset of epilepsy [2], and one of the risk factors for psychosis is a higher seizure frequency [3]. An electroencephalogram (EEG) is a useful tool for research in epilepsy [4,5]. The methods for investigating EEG as discrete signals include linear and nonlinear approaches.…”

Section: Introductionmentioning

confidence: 99%

Linear and non-linear EEG analyses before and after psychosis in patients with epilepsy

Azuma

Nagata

Akechi

2019

E&S

View full text Add to dashboard Cite

Classification of epileptic EEG recordings using signal transforms and convolutional neural networks

Cited by 135 publications

References 47 publications

Real-Time Localization of Epileptogenic Foci EEG Signals: An FPGA-Based Implementation

Real-Time Localization of Epileptogenic Foci EEG Signals: An FPGA-Based Implementation

Fault Diagnosis Based on the Optimization of Characteristic Parameters and Neural Networks of Gearboxes

Linear and non-linear EEG analyses before and after psychosis in patients with epilepsy

Contact Info

Product

Resources

About