1. To understand and analyze common artifacts present in an EEG dataset.
2. To understand various filtering methods for artifact removal from biosignals.
Advanced research in neuroscience investigated the functional dynamics of brain through various noninvasive brain imaging techniques. The neuroimaging techniques includes Electroencephalograph (EEG), functional near-infrared spectroscopy (fNIRS), magnetoencephalography (EMG) that measures the bioelectrical signals generated from living beings. EEG is a portable neuroimaging system that assess different functional brain states. It denotes the neurophysiologic recordings of brains spontaneous electrical activity by placing electrodes on the brain scalp. Electroencephalographic measurements are important for identifying brain activity and behavior. It monitors the voltage fluctuations resulting from ionic current within the neurons. This non-invasive method of brain signal monitoring system was widely used in many areas of clinical work and research, even though one of the main challenging factors in using EEG is its very small signal-to-noise ratio. The characteristic feature of EEG signal is it have a high temporal resolution and easily gets mixed with other biological signals, called artifacts. These unwanted electrical signals that arise from various sources other than the brain, can interfere with the characteristics of neurological phenomena and may negatively use as the source of control in different brain-computer (BCI) systems. Common artefacts are power line noise, muscle contraction, heart activity and eye movement etc. The computational analysis of raw EEG signal from the EEG recordings becomes more complex due to the presence of artifact signals. The unwanted signals have to be eliminated to extract original brain signals and generate a correct EEG data for analysis.
Common Artifacts in EEG signals: The most common signal artifacts appeared during EEG data acquisition is due to different causes, like as dislocated electrode position, oily and not clean hairy leather, blinking of eyes, unwanted muscle movements, line interference electrodes impedance, etc.
Artifacts are mainly classified into two categories: physiological/biological or non-physiological artifacts.
Physiological Artifacts: Physiological artifacts are unwanted bio signals from a physiological origin and is the most complicated artifacts to remove during the analysis. The most significant sources of physiological artifacts are cardiac, pulse, respiratory, sweat, glossokinetic, eye movement (blink, lateral rectus spikes from lateral eye movement), and muscle and movement artefacts(Fig.1).T
Fig 1: Pictorial representation of pulse artifacts(A) and Chewing Artifacts(B) in a window
Non-Physiological Artifacts: Also known as technical artifacts arise from outside the body ie from equipment and environment etc. Main source of extra physiologic artifacts are electronic gadgets and transmission-lines. Common non-physiological artefacts are transmission line artifacts, phone artifacts, electrode artifacts, lead movement artifacts, physical movement artifacts. (Fig.2)
Fig 2: Common artefacts in an EEG signal.
Removal of Artifacts from Bi signals
Bioelectric signals which are low in amplitude and low frequency are stochastic and non-stationary in nature, that means their values are time-dependent and their statistics vary over different points of time. EEG is a time dependent methodology which evaluate and monitor different neurological conditions of a biological subject over some duration of time were further investigated and compared for diagnosing disease conditions. Different methods were proposed to minimize the presence of artifacts after EEG acquisition like blind source artifact separation methods, visual inspection of EEG data and manual deletion of artifactual data segments. Computational analysis of EEG data includes a digital filtration method, that means a mathematical algorithm that operates on a digital dataset in order extract information of interest and remove any unwanted information. Filtering processes are mainly done to retain the frequency component of the specific signal. The preprocessing of EEG signals is mainly done by applying a high-pass filter to filter out slow frequencies less than 0.1 Hz or 1 Hz and a low-pass filter to filter out frequencies above 40 or 50 Hz. Four basic filtering types.
• Low-pass filter – A defined frequency range of EEG data were fixed and all frequencies below the defined frequency are passed and all frequencies above this limit are rejected.
• Band-pass filter – In this filtration method too, a defined frequency was fixed and all frequencies between defined lower and upper frequency limits are passed.
• High-pass – the inverse of the low-pass filter in which all frequencies above a defined frequency limit are passed and all below are rejected.
• Band-stop – often referred to as a « notch filter » is the inverse of the band-pass filter; all frequencies between a defined lower and upper frequency limit are rejected.
Filters designed as, the high-pass filter is derived from the low-pass filter and the band-stop from the band-pass. (Fig.3) Low pass filter retains the high frequency and put through the low frequency according to the parameter settings.
Fig 3: Frequency response of each filtering types.
Image source: https://blricrex.hypotheses.org/filtering-introduction
Also, another proposed filtration method is Finite Impulse Response (FIR) or Infinite Impulse Response (IIR). This digital filtration method eliminates unwanted noises from a raw EEG data. An impulse response defines the time domain-based filtration of a unit impulse signals from a multichannel recording. An FIR filter produces a finite duration of impulse response, after which the output becomes zero and produces equal delays at all frequencies, known as linear phase response. In contrast, IIR filters or recursive filters have an infinite impulse response in which a segment of filter output is used as feedback. This produces unequal delays at different frequencies ie nonlinear phase characteristics. This means that the output signal is shifted in time with respect to the input with some frequency components. IIR filters were widely applied for EEG data filtration since they are computationally more efficient.
Independent Component Analysis (ICA)
Independent component analysis is one of the most important technique applied for electroencephalographic (EEG) signal decomposition and analysis. The main function of this signal preprocessing technique is to find a linear representation of non-Gaussian data whose elements are statistically independent. I.e it minimizes the statistical dependence between the components involved in the signal by including a linear transformation in the random vector. Mostly ocular artefacts and movement-related artifacts from raw EEG data were filtered out by ICA. Advanced Computational application of signal processing techniques includes many algorithms and random vectors that identifies and remove unwanted noises from the bio signal recordings. Clinically, Independent component analysis application to multichannel EEG recording removes unwanted noises and extract the exact human brain signals which computationally analyzed to diagnose diseases without any misleading. It also analysis the feedback associated with bio signal-based systems, such as for Brain-Computer Interface (BCI) and neural prostheses.