The Effects of Acute Moderate Exercise on Cognitive Functioning — Part 7

A Visual Comparison of Analysis Methodologies: RIDE, ICA, & Average ERP Waveforms

To understand the effects of acute moderate exercise on cognitive functioning using ERPs the data needs to be “cleaned” using an analysis methodology. Parts 4, 5, & 6, highlights some of the advantages and disadvantages of utilizing different analysis methodologies to “clean” the data. In short, the analysis methodology will effect the resulting average ERP waveform that is generated for each individual.

At this stage, we can examine the visualizations built earlier to draw some preliminary conclusions of how each analysis methodology affects the average ERP waveform.

Methodological Notes

Average ERP: The process of removing artifacts from the ERP data was done manually by examining individual epochs and comparing them to known ERP artifacts illustrated by an expert in the field.

Average ERP with ICA: ICA was performed on the raw ERP data, and artifact components and individual epochs were removed from the data. In contrast to the artifact removal of the average ERP, fewer epochs were removed from the data due to the use of ICA to enhance the data quality.

Average ERP with ICA & RIDE: The RIDE algorithm was applied the ERP data cleaned using ICA.

Visually comparing methodologies effects on the average ERP waveform

Starting with the grand average waveforms the shape of the peak around ~600 ms appears to differ between the ICA methodology and the RIDE methodology. Recall that the purpose of RIDE is to align individual trials peaks to the median latency before averaging. Doing so will reduce the effect that peak latency variability has when averaging (for more details see part 6).

Figure 1: Grand average ERP waveforms for the rest condition. Both pre and post-rest data are plotted after being manipulated by each methodology.

In figure 1, it is evident that the latency of the peak (P3) around ~600 ms is shifted to the right after being manipulated by the RIDE algorithm. Whether the difference in peak latency is significant between methodologies is unclear. However, this visual comparison provides some indication the methodology does influence the resulting grand averaged ERP waveforms.

Figure 2: Grand average ERP waveforms for the exercise condition. Both pre and post-exercise data are plotted after being manipulated by each methodology.

Again, we see similar changes as in figure 1. The RIDE methodology appears to shift the timing of the “P3” peak to the right.

Figure 3: One individual’s average ERP waveforms for the rest condition. Both pre and post-rest data are plotted after being manipulated by each methodology.

Figure 3 highlights how RIDE can enhance the peak amplitudes within the waveform when individual trials peaks have been aligned to the respective median latency.

Figure 4: One individual’s average ERP waveforms for the exercise condition. Both pre and post-exercise data are plotted after being manipulated by each methodology.

Similarly to figure 3, peak amplitudes are enhanced by RIDE.

Conclusion

From a visual perspective it appears that analysis methodologies can influence the ERP waveform regardless of whether it is at the individual level or across individuals. Although this has no statistical backing, it indicates that the type of analysis methodology can play a role in changing the average ERP waveform. Further analysis will be conducted to illustrate this concept in greater detail in the next section.

The Effects of Acute Moderate Exercise on Cognitive Functioning — Part 6