Adaptive pattern recognition in the analysis of cardiotocographic records

Area of Science:

• Medical technology
• Artificial intelligence in healthcare
• Fetal monitoring

Background:

• Obstetricians manually analyze cardiotocograms to assess fetal state by recognizing accelerative and decelerative patterns in fetal heart rate (FHR).
• Current methods rely on subjective interpretation, necessitating more objective and automated approaches.

Purpose of the Study:

• To develop and validate an artificial neural network-based system for automated recognition of FHR patterns.
• To enhance the system's robustness by incorporating Principal Component Analysis (PCA) and multiresolution techniques.

Main Methods:

• Development of a Multilayer Perceptron (MLP) artificial neural network system.
• Integration of Principal Component Analysis (PCA) to achieve baseline independence of the FHR signal.
• Application of multiresolution techniques within PCA to address identified system limitations, creating the MR-PCA system.
• Validation of the PCA-MLP and MR-PCA systems against assessments from three clinical experts.

Main Results:

• The PCA-MLP and MR-PCA systems demonstrated the ability to recognize accelerative and decelerative FHR patterns.
• The incorporation of PCA and multiresolution techniques improved system performance and robustness.
• Comparative analysis showed the system's potential to assist clinical experts in FHR interpretation.

Conclusions:

• The developed PCA-MLP and MR-PCA systems represent a promising advancement in automated FHR pattern analysis.
• These AI-driven approaches can aid obstetricians in objectively assessing fetal well-being from cardiotocograms.
• Further validation and integration into clinical practice could enhance the accuracy and efficiency of fetal monitoring.