Simultaneous EEG and fMRI experiments require careful set up and pilot testing. Here we provide an overview of how to make sure you obtain best data quality.
The most widely used EEG equipment for simultaneous EEG/BOLD fMRI acquisition consists of the BrainAmp MR plus, BrainVision RecView and BrainVision Analyzer 2. Thus, the Brain Products’ Technical Support team has great experience in helping our customers to solve technical problems concerning EEG/fMRI studies. In this article, we share our experience by presenting an overview of the most important set up steps for a successful EEG/fMRI experiment.
We have seen now and then studies where support was a particular challenge and the scientific outcome was not as good as expected. In all of these problem cases one or more of the following crucial conditions were neglected:
- Careful experiment implementation and porting of the experiment to the MR-environment
- Pilot testing at multiple stages of setting up the experiment
- Ongoing data quality control as the study progresses
- Timely contact with the technical support team.
While many review papers on combined EEG/fMRI address particular theoretical aspects of this methodology, we aim to provide a practical overview on planning and implementing combined EEG/fMRI studies based on our customer support experience in the past years.
Successful combined EEG/fMRI studies are always the result of systematic, step by step implementation and ongoing data quality control during the entire study. Ideally the implementation is an iterative process which aims to establish a setup maximizing signal features of interest while minimizing artifacts and noise in a way that offline MR correction becomes as simple as possible.
We therefore summarize here an optimal workflow for avoiding these pitfalls.
The implementation should be seen as fixed sequence of three consecutive steps which result in the final study setup.
Stage 1: Establish paradigm and EEG measurement outside of the MR environment
The experimental design needs to be established under non-MR conditions in a way that the EEG features of interest are reliably obtained.
Stage 2: Porting the experiment to the MR environment and piloting in the static field (B0)
Within this stage there are a number of aspects to consider in relation to the scanner environment that fall broadly into three categories: volunteer, paradigm, magnetic field specific artifacts.
The scanner bore is in many regards different from conditions in the EEG-laboratory and the potential effect of this on your study volunteers should not be underestimated. The following aspects should be considered with respect to their effects on your volunteers’ comfort and also their ability to perform the tasks you set for them: Stress induced by the unfamiliar, narrow environment, body posture and distractions from the environment (acoustic noise and discomfort, also often reduce alertness).
Stimulus presentation modalities are likely to be different to those used in the lab environment. For example, acoustic stimuli will be delivered using MR compatible headphones/earphones which are different regarding loudness and sound quality. Visual stimuli will be delivered via a different screen, located at some distance from the volunteer and may also be viewed via mirror mounted on the headcoil. Visibility, luminance, contrast, viewing angle, all need to be considered. Importantly the temporal precision of stimulation systems is crucial in EEG experiments and is more critical than for pure fMRI experiments, thus the precision of stimulation event codes (triggers) with regard to the actual stimulation needs to be verified with appropriate methods. Previous Press Release articles on Brain Products StimTrak and Photo Sensor demonstrate the principles.
Magnetic field specific artifacts:
The presence of the static field adds relevant technical noise to all EEG measurements. Subject movements (i.e. heart synchronous head movements, involuntary head movements) and vibrations of technical origin (floor, perhaps Helium-pump or bore air condition) translate via Faradays law into electrical noise. Electrical fields from dimmable scanner room lighting, or in rare cases, mains noise might add electrical noise directly to the EEG. Ideally the noise profile of the individual scanner environment should be assessed by means of a dummy experiment before the first subject is measured. The cardioballistic artifact should be inspected and its removal tested before starting imaging.
Already at this stage, thorough implementation is important: Careful head immobilization is the most important way to reduce the impact of the cardioballistic artifact on EEG. For the cardioballistic artifact correction is an ECG with clearly distinguishable R wave vital. The length of the physical signal path between EEG cap and amplifiers must be restricted to the correct distance between the magnet isocenter and amplifier. Cable swinging (vibrations) and propagated cardioballistic head movements have to be avoided by means of weighting with sand bags and tapes. Routing of the cables must be straight and in line with the central Z-axis to minimize noise further. Depending on the EEG frequency band of interest and noise profile of the scanner, the helium compressor and ventilation systems need to be considered and where necessary switched off during EEG data acquisition.
Pilot tests in the static field give important insights into what needs to be optimized for the final experiment implementation during functional imaging. It is important that the full experiment is carried out: Only by running all phases of the experiment weaknesses will be discovered. For example, sometimes posture turns out to be too uncomfortable over the full duration of the experiment, involuntary head co-movements in response paradigms can add artifacts time locked to the paradigm, or the MR headset can induce artifacts in the EEG. There are a huge number of unexpected side effects which need to be identified and ruled out during pilot testing. Subtle inspection of the raw data for artifacts and correct handling of cardioballistic artifacts must be followed by performing the complete data analysis pipeline for extracting the feature of interest.In addition, prior to measuring any data, make sure you have the correct workspace settings in Recorder for the BrainAmp MR.
Notice that we have not yet run one single BOLD sequence. Taking care of the pilot testing steps listed above can identify problems critical to data quality before we even start scanning. If we are familiar with what our data looks like in the static field we know what data quality we can expect after correction of the gradient artifact.
Once the non imaging related environmental factors have been assessed and the performance of the stimulation (and response) system has been verified, a first volunteer pilot comes into consideration.
Stage 3: Piloting EEG and BOLD imaging simultaneously
If the pilot tests in the static field provide reliably good data quality, EEG with concurrent scanning is the final step of the pilot testing. In contrast to the tests in the static field, gradient fields and RF pulses require particular attention regarding patient and equipment safety. Only the Brain Products released BOLD sequences must be used, cable paths must be straight and centered, only EEG compatible headcoils must be used. All electrodes need to be well prepared to obtain low impedances.
For the pilot tests in the static field, only the stimulation system and EEG-system were interfaced. For the first EEG/fMRI pilot, there are now two more interfaces between scanner gradient system and EEG system needed to enable highest EEG data quality and a simple straight forward gradient correction procedure. The scanner gradient system and BrainAmp MR plus interface on two distinct levels:
(1) Scanner clock and BrainAmp MR plus clock via SyncBox (synchronization level) to ensure a phase locked EEG acquisition in relation to the switching of the gradients during the MR acquisition.
(2) Gradient system and BrainAmp MR plus trigger port (volume- or slice trigger level). The TR needs to be completely stable over time on data point precision and volume/slice triggers need to be sent accurately. Moreover, slice wise gradient correction is facilitated if the following conditions are met: TR (volume) and TR (slice) are integer multiples of the EEG sampling interval (200 µs @ 5kHz). Phantom testing provides an insight about correct function of interfaces and the appropriateness of the sequence timing. However, the timing of the stimuli in relation to the volume acquisition should also be considered. It is important to avoid stimulation repetitions time locked to the TR. A volume artifact based gradient artifact correction would in this caseextinguish all event related EEG patterns.
Once you have successfully performed the simultaneous EEG and BOLD fMRI measurement you should correct the data for the gradient artifact. You should then check that the removal of gradient artifact was successful and that the data quality is the same as the best data that you recorded in the static field. The final steps will be to correct for CB artifact and to perform the whole data analysis pipeline to retrieve the feature of interest.
Only when you are happy that all of the pilot testing steps have been performed satisfactorily, are you ready to start your simultaneous EEG-fMRI study .
While the implementation is an iterative process, maintenance of optimal data quality is a feedback based process. This starts already during the recording when the EEG is corrected online for gradient and pulse artifacts so that the experimenter can rate the data quality online and intervene if needed. We also advise that you analyze your data after each acquisition throughout the study. If the data analyses start only after the last study subject, acquisition problems remain unaddressed, and often cannot be fixed in the offline analysis and precious time and resources are irretrievably lost.
We hope this article might serve as a guideline for successful EEG/fMRI studies and helps to prevent common pitfalls. However, if you still meet problems, need more detailed advice or have problems with data correction, Brain Products’ Technical Support Team will be happy to help you. By asking for our help during the set up and pilot testing phase of your experiment you can save yourself lost time and lost data.