Based On The Measurement Of Accurate Delay Estimation Of Eeg Resonance Gradient Artifact Removal

Recent years with the progress in hardware and software, more and more groups use simultaneous recording of EEG and fMRI to study human brain and nervous system specific illness. By using simultaneous recording of EEG and fMRI , one may want to achieve both high temporal and spatial resolution of brain function. This simultaneous recording technology may become a new tool in brain and brain specific illness research. But when MRI and EEG recording systerm work at the same time ,these two systerm may give artifacts in each other. Due to MRI magnetic field, the recording EEG signals in such enviroment contain very strong MRI artifacts. Before futher process , we should remove the artifacts first. There are two kinds of MRI environment artifact. The first kind is the gradient (or imaging) artifact caused by the switching of the magnetic field gradients. The second kind is the so called ballistocardiographic (BCG) artifact, which is caused by heart-related blood and electrode movements inside the static magnetic field of the MRI scanner, regardless of whether or not MR scanning is being performed. Untill now, several artefact correction methods have been introduced to remove the gradient artifacts. One major methods was firstly proposed by Allen et.al, known as"artifact-template subtraction". This approach assumes that the shape of gradient artifacts is constant over time and additive to the physiological signal. The artifact template computed by averaging several artifact epochs is subsequently subtracted from the artifact-afflicted EEG. Because of this technology cannot remove the artifacts completely, there are many variance technologys based on artifact-template subtraction. these variance methods use ANC,PCA,ICA。Besides artifact-template subtraction,other methods using frequency domain filter, adaptive FIR.In this paper, we first analyse the artifacts produced by most common sequence used in fMRI—EPI sequence and find one signal segment may be fitted by sine waves.Then we propose a method to estimate time delay between the first sample time and the real slice start time by using a sine or cosine function to fit this signal segment. After the delay time have been estimated, we can use this delay to rearrange several successive slice signal to produce a precise artifact in one slice time. At last we use fourier series to fit the rearrange artifacts in one slice, and give a precise mathematical model for the artifacts of one slice. After construct a precise artifacts, we can remove the gradient artifacts by simply subtract it for EEG signal. This method may not need complex postprocessing to remove residual artifacts due to precise delay and curve fitting donot contain much error compare to interpolate.We also introduce another methods based on Allen's, which first use a low pass filter to preprocess the original data, then using the number of sample data points in one slice to produce artifact in one slice and using this artifact as ANC's reference signal to remove it from EEG signal. In the end of this paper we introduce a new method to construct gradien artifacts by using three coils which are prependicular to each other.