Intelligent Neutrosophic Diagnostic System for Cardiotocography Data.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence
• Data Science

Background:

• Cardiotocography (CTG) data uncertainty poses challenges for accurate fetal heart rate classification.
• Machine learning algorithms are crucial for developing efficient diagnostic tools for fetal well-being.

Purpose of the Study:

• To propose and evaluate a novel Interval Neutrosophic Rough Neural Network (INRNN) framework for CTG data classification.
• To enhance the performance of rough neural networks using neutrosophic set theory.

Main Methods:

• Development of an INRNN framework utilizing the backpropagation algorithm.
• Application of WEKA software for comparative analysis of classification algorithms.
• Utilizing boxplots for data visualization and receiver operation characteristic (ROC) curves for performance evaluation.

Main Results:

• The proposed INRNN framework achieved high performance metrics: 95.1% accuracy, 94.95% precision, 95.2% recall, and 95.1% F1-score.
• ROC curve analysis showed high areas under the curve (0.93 for pathologic, 0.90 for normal, 0.85 for suspicious states).
• The INRNN framework demonstrated superior performance compared to traditional neural networks, decision tables, and nearest neighbor algorithms.

Conclusions:

• The proposed neutrosophic diagnostic system is a feasible and efficient classifier for CTG data.
• The INRNN framework offers significant improvements over existing methods for fetal heart rate analysis.
• Future work could involve feature selection to further enhance performance and explore applications in other domains like medical image analysis and social media classification.