Details about the Fsp

Fsp is a measure of response reproducibility, or a signal-to-noise ratio (it is similar to a SNR^2 measure). It is a measure of the variance across the entire averaged response divided by the variance of a single point from sweep to sweep. That single point is usually chosen around the location of Wave V in normal subjects at supra-threshold levels (5-6 ms).

Simplified formula is this:

Fsp = VarS * n / VarSP

  VarS - variance of the response (averaged ABR)
  VarSP - variance for a single point across all sweeps (or, more generally, estimation of residual noise)
  n - number of sweeps averaged together

Fsp threshold was found in a series of experiments (by Claus Elberling) to be 3.1: this means that a reproducible response is present with a 99% probability. But the qualified professional trained to analyze ABR waveforms is still required, because it could happen that it is some noise which happens to synchronize with stimulation rate, for example (electrical interference harmonics) or a muscle artifact/vestibular response.

Also, the 3.1 threshold for Fsp is only valid when residual noise is estimated according to methods 2 and 3. For simple A-B approach it can't be used (or requires adaptation)!

Fsp = F-statistic for single point; Fmp = F-statistic for multiple points
Fmp is an upgraded version of Fsp. Principle is the same, but it is calculated in multiple points and is therefore more accurate.

Fsp was first described in a research paper (and detailed in many others by Claus Elberling):

• Don, M., Elberling, C., and Waring, M.W. (1984). "Objective detection of Averaged auditory brainstem responses, Scand.Audiol. 13, 219-228.