Breathing patterns recognition: A functional data analysis approach.

Area of Science:

• Respiratory physiology
• Biomedical signal processing
• Data science in healthcare

Background:

• Assessing respiratory functionality is vital, especially during pandemics.
• Breathing pattern measurement (tidal volume, respiratory rate) is feasible even in uncooperative patients.
• Current analysis methods are operator-dependent, requiring subjective choices.

Purpose of the Study:

• To develop a semi-automatic, robust, and reproducible procedure for analyzing respiratory data.
• To identify representative breath curves and breathing patterns using Functional Data Analysis (FDA).
• To overcome limitations of operator-dependent analysis in breathing pattern assessment.

Main Methods:

• Utilized Functional Data Analysis (FDA) techniques for respiratory track analysis.
• Employed a three-step process: breath separation, functional outlier detection, and breath clustering (K-medoids with Alignment).
• Validated the method on simulated data and applied it to diverse clinical scenarios.

Main Results:

• Achieved <5% error in minima detection and 99% accuracy in outlier removal.
• Identified five distinct breathing pattern clusters during incremental exercise.
• Successfully differentiated breathing patterns in mechanical ventilation, paradoxical breathing, and age-related changes.

Conclusions:

• A validated, automatic breathing pattern identification algorithm was developed.
• The algorithm extracts representative curves for objective clinical comparison.
• The method demonstrates significant clinical translational value, aligning with physiological conditions.