Application of least-squares minimization and PCA to the optimization of body surface potential maps derived from the standard precordial leads

Brian Lucian LaRocque, University of Texas at El Paso

Abstract

Clinical studies of cardiac patients may benefit from electrocardiogram (ECG) measurements obtained using a multi-lead system consisting of a large number of electrodes arranged in an array encompassing the patient's torso. Two methods have been presented by Lux et al. [14], [15] for optimizing the measurement and ensuing analysis of body surface potential map (BSPM) data. First, an optimal subset of electrodes may be found such that this set contains sufficient signal power to effect an accurate reconstruction of the original BSPM. Second, the original spatial data may be represented by a set of principal components found via principal component analysis. These methods were applied to SCIRun BSPM data derived from the measured standard six precordial leads. Results of the first method showed that a subset of 30 electrodes contains 99.99995% of the trace of the estimated covariance matrix and the original BSPMs may be reconstituted with an average error of 0.15111μVrms with 99.99989% correlation. Applying the optimal transformation to simulated BSPM data from of five patients not used in the estimation of the covariance matrix gave an average error of 0.17659μVrms with 99.99995% correlation. The latter method was used to identify 12 spatial principal components. The associated eigenvectors were used to reconstruct the original BSPM data with an average error of 0.27291μVrms with 99.99971% correlation. Applying these eigenvectors to the test set gave an average error of 0.23255μVrms with 99.99990% correlation. ^

Subject Area

Engineering, Biomedical

Recommended Citation

LaRocque, Brian Lucian, "Application of least-squares minimization and PCA to the optimization of body surface potential maps derived from the standard precordial leads" (2007). ETD Collection for University of Texas, El Paso. AAI1444077.
http://digitalcommons.utep.edu/dissertations/AAI1444077

Share

COinS