7: Inspecting the EEG and Interpolating Bad Channels
Opening the book…
Learning Objectives
In this chapter, you will learn to:
Inspect the raw data to determine what problems exist that might require intervention
Determine whether an EEG channel is “bad” and replace the bad channel with interpolated values
Identify “by eye” the most common types of artifacts, including blinks, saccadic eye movements, and muscle noise
I often say that it’s a miracle that we can record meaningful brain activity from electrodes on the skin, with a thick skull between our electrodes and the brain. Although the EEG is miraculous, it's also an imperfect measure of brain activity. There are many types of artifacts that can contaminate the data, and some of the electrodes may have poor electrical connections to the scalp, creating bad channels.
Several preprocessing steps are necessary prior to averaging so that artifacts and bad channels don’t lead you to draw incorrect conclusions from the ERPs. In particular, we can replace the data from bad channels by interpolation from the good channels, and we can apply artifact rejection and correction to minimize the impact of artifacts.
Prior to performing these steps, it is extremely important that you first perform a visual inspection of a given participant’s EEG so that you understand what kinds of problems are present in that dataset. You also need to spend a lot of time looking at raw EEG data when you are learning to perform preprocessing so that you learn how to identify the most common types of problems.
In this chapter, you will see how to visually inspect EEG data and how to identify bad channels and common artifacts. You’ll also learn how to replace bad channels with interpolated data. The subsequent chapters will cover how to reject and correct artifacts. We’ll look at data from one participant in the mismatch negativity (MMN) experiment from the ERP CORE (Kappenman et al., 2021).
Before we get started, it's important to think about what our goal is for these steps. I find that most people haven’t thought clearly about what they are trying to achieve in their preprocessing pipeline, and this leads to pipelines that are less than ideal.
So, what is the goal of EEG preprocessing? It’s simple: our fundamental goal is to accurately answer the scientific question that the experiment was designed to address. If we include bad channels or large artifacts, this will likely move us away from this goal. However, if we throw out a lot of trials or participants, this will also tend to move us away from our goal by reducing our statistical power. Most preprocessing steps—such as interpolating bad channels or rejecting trials with artifacts—have both costs and benefits, and we need to ask whether the benefits of a given processing step outweigh the costs in terms of answering the scientific question that the experiment was designed to address. We will discuss how to assess these costs and benefits as we go through this chapter and the next two chapters.