Fetal Health Diagnosis Based on Adaptive Dynamic Weighting with Main-Auxiliary Correction Network.

Area of Science:

• Medical Informatics
• Public Health
• Biomedical Engineering

Background:

• Maternal and child health is a global public health priority.
• Fetal cardiotocography (CTG) is crucial for monitoring fetal well-being during pregnancy.
• Category imbalance in CTG data poses challenges for traditional diagnostic models, risking misdiagnosis.

Purpose of the Study:

• To develop an innovative method to improve the classification accuracy of fetal CTG data.
• To address the significant challenge of category imbalance in CTG datasets.
• To enhance early diagnosis and intervention effectiveness for fetal health.

Main Methods:

• Designed a method for adaptive adjustment of misclassification loss function weights (MAAL) to focus on underrepresented samples.
• Developed a primary and secondary correction network model (MAC-NET) for re-evaluating misclassified samples.
• Utilized the UCI publicly available fetal health dataset for validation.

Main Results:

• Achieved 99.39% accuracy on the UCI fetal health dataset.
• Demonstrated excellent performance on other imbalanced domain datasets.
• Effectively alleviated the problem of category imbalance in CTG data.

Conclusions:

• The proposed MAAL and MAC-NET model significantly improves CTG classification accuracy.
• The method offers a robust solution for handling imbalanced data in fetal health monitoring.
• The model exhibits high clinical utility for improving maternal and child health outcomes.