Fourier Analyticstime domain data transformed to the frequency domain
Fourier Transform Analytics
The Secret Sauce behind EvokeDx
icVEP 10 Hz Frequency
The novel isolated-check stimulus pattern varies in contrast, sinusoidally, at 10 Hz. This low-contrast, high temporal frequency pattern is thought to target the M-Cell pathway.
Very Short Run time-domain data collected
10 very short runs (of only a few seconds each) are averaged to create a complex waveform which includes a spectrum of background brain activity (”noise”) obscuring the 10 Hz response.
Fourier Transform of Collected Data from Time to Frequency Domain
The averaged complex waveform is converted by EvokeDx to the frequency-domain using a Fourier transform to deconstruct the complex waveform into its frequency components.
The Frequency Components
From the spectrum of frequencies now revealed, EvokeDx selects the 10 Hz “Response” corresponding to the 10 Hz stimulus, ignoring all irrelevant frequencies.
The 10 Hz Response
Corresponding to the 10 Hz stimulus is isolated.
The 10 Hz response is tested statistically with sophisticated techniques, including “Coherence” (MSC) to assess if the response is significant (unaffected), how significant, or not significant (loss of function).
Fourier Reconstructed Waveform
After selective filtering in the frequency domain, the waveform is then reconstructed into a modified time-domain waveform.
This robust method can have spectacular results in uncovering significant data that is over-powered / obscured by other frequencies such as line-noise from the power grid or other devices in the room.
A measure of Coherence is used to quantify the level of signal power in the response compared to the total power (signal + noise). Termed Magnitude Squared Coherence (MSC), a frequency component of interest is calculated and compared to a critical value and is then displayed graphically to show if a significant response exists at that frequency. Perfect coherence of the recorded response to the stimulus would be indicated by an MSC value of 1.0 (which does not occur in nature).
Additionally, average MSC values for bands of frequency components are calculated that have experimentally been found to reflect distinct neural mechanisms.
This multivariate statistic is calculated on the sine and cosine coefficients of a VEP frequency component to estimate the variability (noise) in the set of responses at the test’s Frequency Component of Interest (FCI). Each individual run’s response component is plotted, with the vector-mean (dot), and a noise circle (radius r) indicating the 95% confidence circle (CC). If the noise circle includes the origin, the response is not significant
Signal to Noise Ratio is the strength of the recorded signal at the frequency of interest relative to the level of noise at the same frequency. A SNR value below 1 indicates that the noise circle overlaps the origin of the sine-cosine plot and that the response is not significant at the .05 level, and above 1 indicates a significant response.
FSTAT assesses if two sweep VEP/ERG functions are statistically different (.05 level) or not. For example, the FSTAT may be used to determine if fellow eye monocular responses are matched or differ sufficiently to raise concerns about a unilateral condition such as amblyopia.
Instant, intuitive test comparisons
A robust comparison feature allows instant comparisons, as examples:
- compare right and left eyes on same day
- compare same eye from a prior exam to assess progression
- compare how a bright response compares to dark response same pattern
- compare how a small check response compares to large check response
- compare how one patient compares to another patient of known characteristics
On-screen graphical help
On viewing results, a concise walk-through of time domain, frequency domain and Fourier analytics is augmented by one-touch help screens. These provide context-sensitive, richly illustrated explanations of all analytic components for all tests.
Scratching your head to understand your first generation VEP / ERG results? Let us help you with elegantly illustrated statistical tools only available in the “Frequency Domain”.