Accurate identification of oxygen desaturation status in COPD by using classifier ensemble

Area of Science:

• Pulmonary Medicine
• Biomedical Engineering
• Machine Learning in Healthcare

Background:

• Accurate identification of oxygen desaturation (OD) is crucial for diagnosing chronic obstructive pulmonary disease (COPD).
• Exercise-induced oxygen desaturation (EIOD) is a specific OD status in COPD patients requiring precise identification.

Purpose of the Study:

• To develop and validate a novel, effective method for identifying EIOD status in COPD patients.
• To improve the performance of EIOD status identification using advanced computational strategies.

Main Methods:

• A classifier ensemble strategy was employed, integrating multiple base classifiers trained on balanced subsets.
• Five distinct features from SpO2 and pulse time-series data were extracted and combined for discriminative feature representation.
• The AdaBoost Algorithm was utilized for integrating the base classifiers.

Main Results:

• The proposed method achieved a significant Area Under Curve (AUC) value of 0.8532 on 6-min walk test (6MWT) data.
• Comparative analysis demonstrated superior global performance compared to existing methods.
• The approach effectively identified EIOD status in the study cohort.

Conclusions:

• The developed ensemble classifier method is effective for identifying EIOD in COPD patients.
• This technique shows potential to assist in the clinical diagnosis and management of COPD.
• The feature extraction and ensemble learning approach offers a robust solution for EIOD detection.