In a virtual sense, Reveal makes a "visual" examination of the EEG. It assigns a perception value to each event, using a range of 0.1 to 1.0 to indicate a rating of "barely perceptible" to "very perceptible". Like human readers making a visual examination of EEG, overwhelming artifact, montage selection, and the channels examined affect the ability to make this determination, allowing you to apply familiar priorities. The factors that affect detection performance in order of priority are:
- Recording quality.
- Montage selection.
- Channel selection.
- Detection parameters.
Recording Quality
Extremely noisy recordings are difficult for both humans and automated detection systems to evaluate, so EEG labs take measures to obtain the best quality recordings possible (e.g., low impedance, good electrode affixing techniques, properly bundled electrode wires, amplifier head box not near any noise source, etc.). Still, when presented with an unavoidably noisy recording SpikeReview does provide some unique capabilities for extracting information. (More on how to do this later.)
Montage
Correct montage selection is fundamental to getting good results, yet it is often overlooked because traces can be viewed on the recording system in a montage completely different from the one selected for detection. The following example illustrates how montage selection and recording quality can affect detection. Using an amplifier reference (labeled "Org" in this recording) we see overwhelming, continuous artifact at P3, A1, and A2. The continuous artifact is confined to those channels, leaving the other cerebral channels relatively unaffected.
The same page viewed with the Av17 montage (all recorded channels referenced to the average of F7, Fz, F3, F4, F8, T3, C3, Cz, C4, T4, T5, P3, Pz, P4, T6, O1, and O2). Note that the overwhelming artifact at P3 has contaminated all the other channels.
Below is the same page viewed using an A1+A2 reference.
Given the effects of large continuous artifact, why use an average reference?
First, a referential montage offers better topographic mapping (and clustering for SpikeReview) than a bipolar montage. (To understand why bipolar montages are difficult for topographic mapping, consider this question. What value should be plotted at Fp2 on a topographic voltage map when the montage contains both Fp2-F4 and Fp2-F8?)
Secondly, the average reference removes the possibility that an "active reference" is used. For example, you don’t want to use an A1+A2 reference for patients with temporal spikes because this causes maximum deflection away from the spike focus. Thus, SpikeDetector’s default scanning montage, which includes 10-20 electrodes and an Av17 reference, is a good choice when there is no a priori knowledge about the epileptogenic focus. However, if some information is known about the likely focus, e.g., not central, then it may be preferable to use a C3+C4 or Cz reference. Av17 is dependent on more electrodes, but it can probably handle a mildly poor connection on a single electrode (unlike the case shown above) better than other references that use fewer electrodes.
Another, option is to use the predefined Av12 reference, which uses fewer electrodes (F7, F8, Fz, Cz and Pz are eliminated) than Av17.
If you use an average reference, verify that it includes only the channels you are recording, and that it does not include heavily contaminated channels. To illustrate these points, the page below shows the same data after creating an "Av15" reference in Insight (copied from Av17, however having P3 and P4 removed) where P3 is no longer included in the average reference, and does not contaminate the resultant montage.
UPDATE: Reveal version 2002.11.19 and later: The spike detection algorithm has been updated, and there are two resulting UI changes. First, the “DefaultScalp” protocol now used a Laplacian montage (LaplacianNoFpzOz) rather than Ref-Group with the default reference. This montage has the benefits of both referential (allows determination of scalp sign and topographic maps) and bipolar (one or two bad electrodes do not contaminate all the channels) montages. This is the montage used to mark the vast majority of the training records. Secondly, the Sensitivity (uV/mm) setting can now be set automatically. The computation uses a linear NN, so it should be applicable to any range (including intracranial) of EEG activities. The computation uses the first three minutes of data and is robust with respect to the presence of calibrations signals and artifacts. The automatically determined sensitivity can be checked from the Montage page after processing starts. Currently, the Automatic Sensitivity feature is enabled only on the Montage page for spike detection. If the record being analyzed is shorter than three minutes, then the AutoEnd option must be turned on so that the Automatic Sensitivity calculation uses the existing record length.
