Cross-recurrence analysis for pattern matching of multidimensional physiological signals.

Area of Science:

• Nonlinear dynamics
• Time series analysis
• Biomedical signal processing

Background:

• Cross-recurrence quantification analysis (CRQA) quantifies nonlinear interrelationships in time series.
• CRQA is valuable for data mining but underexplored for multidimensional physiological signals.

Purpose of the Study:

• To develop and evaluate a novel CRQA-based methodology for clustering multidimensional physiological signals.
• To assess the performance of CRQA statistics as dissimilarity measures for patient stratification.

Main Methods:

• Utilized CRQA statistics as dissimilarity measures between pairs of signals.
• Applied clustering to 3D spatiotemporal vectorcardiogram (VCG) signals from healthy and diseased patients.
• Evaluated Lmax and a novel measure, Rτmax, for their clustering performance.

Main Results:

• The Lmax measure achieved clustering that closely matched ground truth patient diagnoses.
• The proposed Rτmax measure demonstrated superior performance in matching diagnoses after signal rescaling.
• CRQA effectively identified patterns in complex spatiotemporal VCG data.

Conclusions:

• CRQA-based dissimilarity measures offer a powerful approach for clustering physiological time series.
• The Lmax and Rτmax metrics show promise for non-invasive patient diagnosis and stratification.
• This methodology advances the application of CRQA in biomedical signal analysis.