Identification of eupneic breathing using machine learning

Area of Science:

• Physiology
• Neuroscience
• Machine Learning

Background:

• The diaphragm muscle (DIAm) is the primary inspiratory muscle.
• Heterogeneity in DIAm electromyography (EMG) activity in awake animals reflects diverse behaviors.
• Accurate identification of breathing-related DIAm EMG is essential but challenging.

Purpose of the Study:

• To develop a reliable and efficient method for classifying eupneic breathing from DIAm EMG in awake animals.
• To address the lack of reproducible strategies for analyzing DIAm EMG during breathing.

Main Methods:

• Utilized hierarchical clustering in a four-dimensional feature space to analyze DIAm EMG data.
• Applied an unsupervised machine learning model to classify breathing patterns.
• Implemented an automated threshold for the clustering dendrogram.

Main Results:

• The developed model successfully identified eupneic breathing with high accuracy (0.88), F1 score (0.92), and specificity (0.70).
• The method proved robust and reliable for distinguishing breathing-related EMG activity.
• Demonstrated scalability and efficiency in DIAm EMG analysis.

Conclusions:

• The unsupervised machine learning approach provides a robust tool for analyzing DIAm EMG.
• This technique enhances the efficiency and reliability of studying the neuromotor control of breathing.
• Minimizes potential bias in DIAm EMG analysis, facilitating research on breathing control.