Orthogonal expansions: their applicability to signal extraction in electrophysiological mapping data

Area of Science:

• Biophysics
• Computational Neuroscience
• Signal Processing

Background:

• Orthogonal expansions, including singular-value decomposition (SVD), Karhunen-Loève transform (KLT), and principal-component analysis (PCA), are mathematical techniques used for data dimensionality reduction and feature extraction.
• These methods rely on the orthogonality of eigenvectors to isolate components within a signal.
• Their application in analyzing complex biological signals, such as electrophysiological data from the heart and brain, requires careful consideration of the underlying source characteristics.

Purpose of the Study:

• To evaluate the suitability of orthogonal expansion techniques for identifying distinct electrophysiological sources in the heart and brain.
• To investigate the relationship between the eigenvectors derived from these expansions and the actual physiological sources generating the signals.
• To determine the validity of interpreting the results of orthogonal expansions in terms of specific physiological events.

Main Methods:

• A current dipole source model was employed to simulate electrophysiological signals.
• Orthogonal expansion techniques, specifically SVD, KLT, and PCA, were applied to these simulated signals.
• The properties of the resulting eigenvectors were analyzed in relation to the known source distributions.

Main Results:

• Orthogonal expansion eigenvectors are designed to extract features of a single source only if all other signals are orthogonal to it.
• While the number of significant eigenvectors can correlate with the number of signal components, a direct one-to-one correspondence with individual sources is not guaranteed.
• Representing even a single, non-stationary source may necessitate a large number of eigenvectors, complicating interpretation.

Conclusions:

• Orthogonal expansions may not accurately isolate individual physiological sources due to potential non-orthogonality of biological signals.
• A direct physiological interpretation of the data derived from these expansions is generally inappropriate.
• The complexity and non-stationary nature of biological signals limit the direct mapping of eigenvectors to specific electrophysiological events.