Automatika (Oct 2018)

Performance analysis of swarm intelligence algorithms in removal of ECG artefact from tainted EEG signal

  • S. Suja Priyadharsini,
  • S. Edward Rajan

DOI
https://doi.org/10.1080/00051144.2018.1541642
Journal volume & issue
Vol. 59, no. 3-4
pp. 408 – 415

Abstract

Read online

Electroencephalogram (EEG) is the recording of electrical activities of the brain. It is contaminated by other biological signals, known as artefacts. In this research paper, the performance analysis of three swarm intelligence incorporated adaptive neuro fuzzy inference system (ANFIS) - based techniques is made with respect to ECG artefact removal from the corrupted EEG signal. Swarm intelligence algorithms such as improved artificial immune system (IAIS), artificial immune system (AIS) and particle swarm optimization (PSO) are employed for artefact removal, by tuning the parameters of ANFIS individually. The performances of the methods are experimentally validated for both simulated and real data sets. Measures such as signal to noise ratio (SNR), mean square error (MSE) value, correlation coefficient, power spectrum density plot, sensitivity, specificity and accuracy are used for analysing the performance of the methods of simulated data set. The sensitivity, specificity and accuracy of ANFIS-tuned IAIS (ANFIS-IAIS), are found to be 94.9%, 100% and 99.2%, respectively The sensitivity, specificity and accuracy of ANFIS-AIS and ANFIS-PSO are 91.9%, 100%, 98.7% and 87.9%, 100%, 98.3%, respectively. From the results, it is found that ANFIS-IAIS is more effective in removing ECG artefacts from EEG signals than ANFIS-AIS and ANFIS-PSO. Analiza vladanja algoritama inteligencije roja u odstranjivanju ECG artefakata iz onečišćenog EEG signala Elektroencefalografija (EEG) je postupak snimanja električnih aktivnosti mozga. Postupak je zagadena drugim biološkim signalima zvanim artefaktima. U radu je predstavljena analiza vladanja tri pristupa zasnovana na inteligenciji roja i pristupu adaptivnog neuro-neizrazitog zaključivanja (ANFIS) za uklanjanje ECG artefakta iz onečišćenog EEG signala. Algoritmi inteligencije roja kao što su unaprijedeni umjetni imunosni sustav (IAIS), umjetni imunosni sustav (AIS) i optimizacija roja čestica (PSO) korišteni su za uklanjanje artefakata individualnim podešavanjem parametara ANFIS pristupa. Vladanje metoda eksperimentalno je potvrdeno na simulacijskim i stvarnim podacima. Mjerila kao što su odnos signal-šum (SNR), srednja kvadratna pogreška (MSE), korelacijski koeficijent, graf gustoće spektra snage (PSD), osjetljivost, specifičnost i točnost korištena su za analizu vladanja metoda na simulacijskim podacima. Osjetljivost, specifičnost i točnost IAIS pristupa uz ANFIS podešavanje parametara (ANFIS-IAIS) pokazalo se kao, redoslijedom, 94,9%, 100%i 99,2%, što je veće u odnosu na druge metode. Osjetljivost, specifičnost i točnost ANFIS-AIS i ANFIS-PSO pristupa je 91,9%, 100% i 98,7% te 87,9%, 100% i98,3%, redoslijedom. Rezultati pokazuju da je ANFIS-IAIS pristup efektniji u otklanjanju ECG artefakata iz EEG signala u odnosu na ANFIS-AIS i ANFIS-PSO pristupe.

Keywords