EEG Micrcontinuity lite

EEG Microcontinuity

The concept of EEG microcontinuity, as described in the original paper by Kemp et al (2000), proposes a new SWA metric which is, citing from the Abstract, "the fraction (0%-100%) of the current slow wave which continues in the near-future (0.02 s later) EEG".

This metric is implemented by translating relevant bits from the C# code published by Marco Roessen, who in turn did the coding of the original concept in collaboration with Bob Kemp.

The algorithm, following the logic and comments in said C# code, proceeds like so:

  1. Perform SU and SS reduction;
  2. Compute PiB value;
  3. Detect artifacts;
  4. Smooth SU and SS;
  5. Detect events;
  6. Re-smooth signals and detecting jumps;
  7. Compute final gains.

In doing my part, I got stuck in the C# thicket just after step 2. Eventually, as part of a debugging effort, I noticed that the intermediate results coming out after step 1, bore remarkable semblance to the ultimate course of "SW%" as it appears on fig. 2 (chart 2).

Step 1, thus, produces the "lite" SW% metric, at each page p, as SS[p] - SU[p], where SS and SU are computed as shown in eq. 22 of the cited paper. Note that these values are computed over the entire page length (typically, 30 sec) rather than the shorter 1-sec intervals used in the paper.

For what it is, the SWA (or SW%) profiles of EEG Microcontinuity do look handsome to me, nicely following the ultradian cycle and expressing the SWA swing with clear emphasis. And it even elicits a novel feature, which is a steadier buildup of the metric towards the end of a slow-wave hump where PSD yields a more steep increase early on into the period.

Please take this with a pinch of salt.

Artifact detection

The SS-SU difference is used for artifact detection.