[A Review on Automatic Detection Algorithm for Patient-Ventilator Asynchrony during Mechanical Ventilation].

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine

Background:

• Patient-ventilator asynchrony (PVA) detection traditionally relied on manual rule-based systems.
• Manual methods show limitations in adaptability and sensitivity to subtle patient status changes.

Approach:

• Explores the evolution of PVA recognition from manual interpretation to machine learning (ML) and deep learning (DL) techniques.
• Highlights ML/DL algorithms like logistic regression, SVM, random forest, HMM, CNNs, and LSTMs for automated PVA detection.
• Discusses the potential of reinforcement learning and self-supervised learning to address data demands of DL methods.

Key Points:

• Machine and deep learning offer more robust and universal PVA detection compared to manual methods.
• Deep learning excels at feature extraction but requires substantial labeled data.
• Cross-dataset validation is crucial for future algorithm development and generalization.

Conclusions:

• AI-driven PVA recognition significantly enhances mechanical ventilation monitoring.
• Future research should focus on data-efficient learning and robust validation strategies.
• Optimizing PVA detection is key to improving patient outcomes during mechanical ventilation.