Classification of impedance cardiography dZ/dt complex subtypes using pattern recognition artificial neural networks

Area of Science:

• Biomedical Engineering
• Cardiovascular Physiology
• Signal Processing

Background:

• Accurate detection of characteristic points in impedance cardiography (ICG) signals, particularly the X point, is essential for calculating hemodynamic parameters like stroke volume (SV) and cardiac output (CO).
• Beat-to-beat calculations of these parameters are often hindered by the variability and complex subtypes within ICG waveforms, impacting detection accuracy.
• Automated classification of ICG complexes is needed to improve the reliability of hemodynamic parameter extraction.

Purpose of the Study:

• To develop and validate an automated method for classifying impedance cardiography (ICG) complex subtypes.
• To support the accurate detection of ICG characteristic points and subsequent extraction of hemodynamic parameters.
• To investigate the efficacy of a novel pattern recognition artificial neural network (PRANN) for ICG waveform classification.

Main Methods:

• Implementation of a novel pattern recognition artificial neural network (PRANN) utilizing a divide-and-conquer strategy.
• Training, testing, and validation of the PRANN on ICG datasets from eight volunteers, using measurements from eight electrodes.
• Validation on independent datasets to assess the generalizability and robustness of the classification method.

Main Results:

• The PRANN achieved an average accuracy of 96% in classifying ICG complex subtypes on the training and testing datasets.
• The network demonstrated strong performance on independent validation datasets, with accuracies of 83% and 80%.
• The results indicate the PRANN's capability to effectively differentiate between various ICG complex waveforms.

Conclusions:

• The developed PRANN is a promising tool for the automated classification of ICG complex subtypes.
• This automated classification facilitates more accurate detection of ICG characteristic points and hemodynamic parameter extraction.
• The PRANN approach holds potential for advancing beat-to-beat hemodynamic monitoring using impedance cardiography.