IEEE Transactions on Signal Processing
Computer Methods and Programs in Biomedicine
| Hi-index | 35.69 |
Measurable magnetoencephalographic responses of the cortex due to an electrical stimulus at the wrist start 20 ms after the stimulus. This early magnetic response is known as the N20m, which can be seen by averaging over hundreds of stimulation epochs. Applying Independent Component Analysis (ICA) based on time-delayed decorrelation to such data allows the extraction of the single responses starting 20 ms after the stimulus without the need for averaging. One of the independent components has a field pattern that is very similar to the N20m. Using this independent component, it is found that the response at 20 ms is stable over a measurement session lasting 4000 s and containing 12 000 stimulations, whereas later responses show highly significant changes over time. To suppress slower activity and noise in the data, a high pass of 55 Hz is applied to the data. One of the subsequently calculated independent components shows that the response at 20 ms is much clearer than before filtering. Analyzing the amplitude distribution of this response yields that 97% of the stimulations have a measurable response above base line level, whereas for conventional methods such as projection and notch filtering, only 91% of the responses are detectable. The high degree of measurable responses indicates the signal separation power of independent component analysis, and it supports the hypothesis that the early stages of sensory cortical processing can be described as a linear processing chain with small variability, at least from a macroscopic point of view.