Artifact reduction in magnetogastrography using fast independent component analysis.

Irimia A, Bradshaw LA
Physiol Meas. 2005 26 (6): 1059-73

PMID: 16311453 · DOI:10.1088/0967-3334/26/6/015

The analysis of magnetogastrographic (MGG) signals has been limited to epochs of data with limited interference from extraneous signal components that are often present and may even dominate MGG data. Such artifacts can be of both biological (cardiac, intestinal and muscular activities, motion artifacts, etc) and non-biological (environmental noise) origin. Conventional methods-such as Butterworth and Tchebyshev filters-can be of great use, but there are many disadvantages associated with them as well as with other typical filtering methods because a large amount of useful biological information can be lost, and there are many trade-offs between various filtering methods. Moreover, conventional filtering cannot always fully address the physicality of the signal-processing problem in terms of extracting specific signals due to particular biological sources of interest such as the stomach, heart and bowel. In this paper, we demonstrate the use of fast independent component analysis (FICA) for the removal of both biological and non-biological artifacts from multi-channel MGG recordings acquired using a superconducting quantum intereference device (SQUID) magnetometer. Specifically, we show that the signal of gastric electrical control activity (ECA) can be isolated from SQUID data as an independent component even in the presence of severe motion, cardiac and respiratory artifacts. The accuracy of the method is analyzed by comparing FICA-extracted versus electrode-measured respiratory signals. It is concluded that, with this method, reliable results may be obtained for a wide array of magnetic recording scenarios.

MeSH Terms (15)

Action Potentials Algorithms Artifacts Diagnosis, Computer-Assisted Electromyography Gastrointestinal Motility Humans Magnetics Muscle, Smooth Muscle Contraction Principal Component Analysis Reproducibility of Results Respiratory Mechanics Sensitivity and Specificity Signal Processing, Computer-Assisted

Connections (1)

This publication is referenced by other Labnodes entities: