UNSUPERVISED CLUSTERING AND ANALYSIS OF CONTRACTION-DEPENDENT FETAL HEART RATE SEGMENTS

Area of Science:

• Medical Informatics
• Signal Processing
• Fetal Monitoring

Background:

• Computer-aided interpretation of fetal heart rate (FHR) and uterine contraction (UC) is crucial for delivery room monitoring but faces challenges.
• Lack of standardized labels and difficulties in timely fetal state prediction hinder current cardiotocography (CTG) analysis.

Purpose of the Study:

• To develop an unsupervised method for understanding UC-dependent FHR responses.
• To provide a complete method for FHR-UC segment clustering and analysis.
• To propose a novel trajectory-based interpretation for FHR signals.

Main Methods:

• Utilized Gaussian process latent variable models and density-based spatial clustering for FHR-UC segment analysis.
• Mapped UC-dependent FHR segments into a visually interpretable space.
• Defined three metrics for FHR trajectory analysis.

Main Results:

• Successfully clustered and analyzed FHR-UC segments in an unsupervised manner.
• Developed a novel method for interpreting FHR trajectories based on UC.
• Demonstrated the method's efficacy using an open-access CTG database.

Conclusions:

• The proposed unsupervised, trajectory-based approach offers a promising alternative to traditional supervised methods for FHR interpretation.
• This method enhances the understanding of FHR-UC dynamics, potentially improving fetal monitoring accuracy.
• The developed techniques provide a foundation for more robust computer-aided interpretation of CTG recordings.