Unsupervised study of plethysmography signals through DTW clustering

Area of Science:

• Respiratory Physiology
• Neurotoxicology
• Bioinformatics

Background:

• Current analysis of plethysmography time series relies on limited descriptors of individual breathing cycles, failing to capture dynamic changes in respiratory behavior.
• Understanding neurotoxin effects on respiration requires methods that account for the variability and temporal evolution of breathing patterns.

Purpose of the Study:

• To develop a novel, automated procedure for analyzing mouse plethysmography signals.
• To introduce a robust method for segmenting respiratory cycles and identifying typical patterns.
• To create a symbolic representation for visualizing and quantifying the temporal evolution of breathing behavior.

Main Methods:

• A new, robust segmentation algorithm for respiratory cycles in plethysmography signals.
• Dynamic Time Warping (DTW)-based clustering to identify reference respiratory cycles.
• Symbolic representation generation by matching new respiratory cycles to reference sequences.

Main Results:

• The method enables accurate extraction of clinically relevant respiratory cycles and ventilation descriptors (e.g., tidal volume, inhalation/exhalation duration).
• A symbolic representation provides a visual and quantitative tool to assess breathing behavior and its evolution over time.
• The approach was successfully applied to plethysmography data from mice with different genotypes exposed to a neurotoxin.

Conclusions:

• The proposed algorithm offers a significant advancement in the analysis of plethysmography data.
• This novel approach facilitates the study of subtle, time-evolving effects of neurotoxins on the respiratory system.
• The symbolic representation enhances the quantitative assessment of respiratory dynamics in preclinical research.