GES 400 MR Artifact Handling
Masking the EEG
Data acquired within the MRI scanner is subject to artifacts above and beyond those traditionally associated with EEG acquisition. Three unique artifacts that we need to deal with are:
Imaging artifacts: During each head volume acquisition, both the gradient switching and radio frequency pulse pattern will produce strong artifacts that will obscure the EEG. However, the TR chosen for a particular scan protocol will provide this artifact with a very specific temporal pattern.
Cardioballistic artifacts: Every time the heart beats, the resultant micromovement and blood flow of the head causes this periodic artifact. Cold head artifacts: The cryogenic pump of the MR system causes a systematic artifact.
Cleaning Up with the GES 400 MR
New with Net Station 5.2!
Pulse artifact detection tool based on Iannotti, et al. (2015).
In this publication in Brain Topography, Iannotti, et al. from Geneva University Hospital introduce a new EEG-based method for detecting the pulse artifact (PA) with significant benefits over the standard ECG-based methods. The method uses the anti-symmetric scalp voltage potential topography of the pulse artifact, captured by the EEG, to identify the artifacts. At least as accurate as ECG-based methods, and often more so, the EEG-based method can replace ECG-based methods, being especially useful in cases where it is difficult to obtain clean ECG data, and can also complement ECG-based artifact detection as a secondary measurement.
Although imaging artifacts initially render the EEG completely unreadable, they do have characteristics that make them ideal targets for artifact correction. The gradient switching and radio frequency pulses occur within nanoseconds of each other across each volume acquisition, thereby creating a consistent artifact waveform profile with respect to each TR. As the MRI scanner sends a TR trigger to the EEG file, there is a clear method of identifying the onset and profile of the artifact. This promotes a relatively straightforward method of artifact correction, calculate an average of the imaging artifact waveform and subtract it from the simultaneously recorded EEG.
While the theory is simple, the challenge comes in proper implementation. Since the EEG acquisition and MRI scanner use separate computers, their clocks are independent. In addition, MRI scanner itself has a very high amplitude and rapid slew rate. Therefore, tiny amounts of jitter between the two clocks will result in the imaging artifact profile varying from one volume to the next. As this inconsistency reduces the efficacy of the average artifact subtraction method, it is not uncommon for additional steps to be introduced to the data cleaning workflow, such as signal interpolation and adaptive noise cancellation.
These limitations have led us to take a novel approach. We have created a situation where the EEG acquisition is effectively running the on the MRI system’s clock. This is done by introducing a phase-locked loop between the MRI and EEG system, which results in the jitter between the two clocks being reduced to a negligible amount (5-10 nanoseconds). In another example of how we closely work with our customers when developing our new products, we are licensing this synchronization technique from Mark Cohen PhD from UCLA’s David Geffen School of Medicine and it is covered by US Patient No. 7,286,871.
Additionally, front-end filters within our FICS (Field Isolation Containment System) unit decrease the amplitude of the artifact prior to signal digitization. By the skillful combination of these two techniques we are then able to effectively remove the imaging artifact using the simple average artifact subtraction method. The procedure can be implemented in real-time during the MR scan for EEG data monitoring and through our Waveform Tools when following your data processing workflow.
Balistocardiogram (BCG) artifacts are difficult to handle and remove for several reasons: 1) within each electrode, their amplitudes and duration vary between successive heartbeats, 2) between electrodes their amplitudes and pattern vary, 3) their energy distribution overlaps with the frequency of the EEG, and 4) they share similar morphology with certain EEG phenomena. There are a handful of techniques that have been proposed for handling BCG artifacts (see Groiller et al., 2007) and we currently recommend the technique of Niazy et al. (2005) and implemented in FMRIB EEGLAB plug-in). We have implemented this method within Net Station to optimize workflow.
Currently we recommend that the MR’s cold-head system be turned off during EEG acquisition. Although this is currently standard practice for simultaneous EEG/fMRI acquisition, we appreciate that it is not an ideal situation. Therefore we are investigating methods for cryo-ballistic artifact correction and we hope to implement these in future release of our software.