A method for detecting high-frequency oscillations using semi-supervised k-means and mean shift clustering

Du, Yuxiao; Sun, Bo; Lu, Renquan; Chun-ling, Zhang; Wu, Hao

doi:10.1016/j.neucom.2019.03.055

Cited by 26 publications

(10 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The conventional method of HFOs detection is usually based on either the band-pass signal or the time-frequency diagram. In terms of signal, single-step detection methods with different characteristics of Teager energy operator, wavelet entropy, fuzzy entropy, short-time energy, and so on have been studied in the past 10 years [ 17 – 24 ]. However, it is difficult for these methods to distinguish HFOs from some artifacts, such as spikes, pulse-like artifacts, and signals with harmonics [ 20 ].…”

Section: Introductionmentioning

confidence: 99%

[Retracted] Automatic Detection of High‐Frequency Oscillations Based on an End‐to‐End Bi‐Branch Neural Network and Clinical Cross‐Validation

Liu

Wei

Wang

et al. 2021

Computational Intelligence and Neuroscience

View full text Add to dashboard Cite

Accurate identification of high-frequency oscillation (HFO) is an important prerequisite for precise localization of epileptic foci and good prognosis of drug-refractory epilepsy. Exploring a high-performance automatic detection method for HFOs can effectively help clinicians reduce the error rate and reduce manpower. Due to the limited analysis perspective and simple model design, it is difficult to meet the requirements of clinical application by the existing methods. Therefore, an end-to-end bi-branch fusion model is proposed to automatically detect HFOs. With the filtered band-pass signal (signal branch) and time-frequency image (TFpic branch) as the input of the model, two backbone networks for deep feature extraction are established, respectively. Specifically, a hybrid model based on ResNet1d and long short-term memory (LSTM) is designed for signal branch, which can focus on both the features in time and space dimension, while a ResNet2d with a Convolutional Block Attention Module (CBAM) is constructed for TFpic branch, by which more attention is paid to useful information of TF images. Then the outputs of two branches are fused to realize end-to-end automatic identification of HFOs. Our method is verified on 5 patients with intractable epilepsy. In intravalidation, the proposed method obtained high sensitivity of 94.62%, specificity of 92.7%, and F1-score of 93.33%, and in cross-validation, our method achieved high sensitivity of 92.00%, specificity of 88.26%, and F1-score of 89.11% on average. The results show that the proposed method outperforms the existing detection paradigms of either single signal or single time-frequency diagram strategy. In addition, the average kappa coefficient of visual analysis and automatic detection results is 0.795. The method shows strong generalization ability and high degree of consistency with the gold standard meanwhile. Therefore, it has great potential to be a clinical assistant tool.

show abstract

Section: Introductionmentioning

confidence: 99%

[Retracted] Automatic Detection of High‐Frequency Oscillations Based on an End‐to‐End Bi‐Branch Neural Network and Clinical Cross‐Validation

Liu

Wei

Wang

et al. 2021

Computational Intelligence and Neuroscience

View full text Add to dashboard Cite

show abstract

“…The classification and differentiation of data with approximate features using mean-shift clustering analysis [36] is a hill-climbing algorithm based on kernel density estimation, which is a clustering algorithm that does not require the number of clusters to be specified. Processing [37] and cluster analysis [38] are the commonly used cluster analysis methods in the field of data mining. The purpose of mean-shift clustering is to find the data sample points of the same cluster along the direction of increasing density.…”

Section: Development and Application Of The Bp Prediction Modelmentioning

confidence: 99%

Calibration and Verification of Dynamic Particle Flow Parameters by the Back-Propagation Neural Network Based on the Genetic Algorithm: Recycled Polyurethane Powder

Fan

Pan

et al. 2019

Materials

View full text Add to dashboard Cite

The discrete element method (DEM) is commonly used to study various powders in motion during transportation, screening, mixing, etc.; this requires several microscopic parameters to characterize the complex mechanical behavior of the particles. Herein, a new discrete element parameter calibration method is proposed to calibrate the ultrafine agglomerated powder (recycled polyurethane powder). Optimal Latin hypercube sampling and virtual simulation experiments were conducted using the commercial DEM software; the microscopic variables included the static friction coefficient between the particles, collision recovery coefficient, Johnson–Kendall–Roberts surface energy, static friction coefficient between the particles and wall, and collision recovery coefficient. A predictive model based on genetic-algorithm-optimized feedforward neural network (back propagation) was developed to calibrate the microscopic DEM simulation parameters. The cycle search algorithm and mean-shift cluster analysis were used to confirm the input parameters’ range by comparing the mean value of the dynamic angle of repose measured via the batch accumulation test. These parameters were verified by the baffle lifting method and the rotating drum method. This calibration method, once successfully developed, will be suitable for use in a variety of fine viscous powder dynamic flow conditions.

show abstract

“…In last decades, HFOs detectors were developed using a long pipeline including artefact rejection, filtering, feature engineering, feature selection and eventually a classification step for false detection rejection [2] Recent studies perform feature extraction using supervised or unsupervised machine learning techniques [3,4]. Most detectors are developed for iEEG, very few methods were proposed to automatically detect HFOs from scalp EEG [5][6][7] and most of them are extensions of iEEG HFOs detectors, are semi supervised and require definition of threshold(s) [5,6] . The only non-threshold based scalp HFOs detector was proposed in [7] using a semi-supervised k-means algorithm followed by a mean shift algorithm to classify suspicious HFOs.…”

Section: Introductionmentioning

confidence: 99%

“…Most detectors are developed for iEEG, very few methods were proposed to automatically detect HFOs from scalp EEG [5][6][7] and most of them are extensions of iEEG HFOs detectors, are semi supervised and require definition of threshold(s) [5,6] . The only non-threshold based scalp HFOs detector was proposed in [7] using a semi-supervised k-means algorithm followed by a mean shift algorithm to classify suspicious HFOs. To overcome all the thresholding and post detector visual reviews, we propose here an automatic classification between HFOs (80-500 Hz) and EEG signal outside this frequency range (non HFOs) based on deep learning.…”

Section: Introductionmentioning

confidence: 99%

1D vs 2D convolutional neural networks for scalp high frequency oscillations identification

MILON-HARNOIS¹,

JRAD²,

Schang³

et al. 2022

ESANN 2022 Proceedings

View full text Add to dashboard Cite

Scalp High Frequency Oscillations (HFOs) are promising biomarkers of epileptogenic zones. Since HFOs visual detection is strenuous, there is a real need to develop accurate HFOs automatic detectors. In this paper, we present a comparative study of two detectors: onedimensional (1D) Convolutional Neural Networks (CNN) running on High-Density Electroencephalograms signals and two dimensional (2D) CNN on time-frequency maps of those signals. Experimental results show that 1D-CNN enables easy end-to-end learning of preprocessing, feature extraction and classification modules while achieving competitive performance.

show abstract

A method for detecting high-frequency oscillations using semi-supervised k-means and mean shift clustering

Cited by 26 publications

References 25 publications

[Retracted] Automatic Detection of High‐Frequency Oscillations Based on an End‐to‐End Bi‐Branch Neural Network and Clinical Cross‐Validation

[Retracted] Automatic Detection of High‐Frequency Oscillations Based on an End‐to‐End Bi‐Branch Neural Network and Clinical Cross‐Validation

Calibration and Verification of Dynamic Particle Flow Parameters by the Back-Propagation Neural Network Based on the Genetic Algorithm: Recycled Polyurethane Powder

1D vs 2D convolutional neural networks for scalp high frequency oscillations identification

Contact Info

Product

Resources

About