AI driven interpretable deep learning based fetal health classification.

Area of Science:

• Obstetrics and Gynecology
• Medical Informatics
• Artificial Intelligence in Medicine

Background:

• Fetal health assessment is critical in obstetrics.
• Current diagnostic methods can be improved for efficiency and effectiveness.
• Deep learning offers potential for enhanced medical diagnostics.

Purpose of the Study:

• To propose a deep learning model for classifying fetal health into Normal, Suspect, and Pathological categories.
• To improve the efficiency and effectiveness of fetal health diagnostic processes.
• To enhance the transparency and clinical adoption of fetal health assessments through explainable AI.

Main Methods:

• A deep neural network (DNN) model was developed for fetal health analysis.
• The model utilized a dataset with 21 attributes from Cardiotocography (CTG) recordings.
• Explainable deep learning techniques, including feature importance and saliency analysis, were employed for model interpretation.

Main Results:

• The proposed DNN model achieved high performance: 0.99 accuracy, 0.93 sensitivity, 0.93 specificity, 0.96 AUC, 0.93 precision, and 0.93 F1-score.
• Comparative analysis showed the DNN model outperformed six baseline models (Logistic Regression, KNN, SVM, Naive Bayes, Random Forest, Gradient Boosting).
• The model demonstrated superior accuracy compared to all baseline methods.

Conclusions:

• Deep learning, specifically the proposed DNN model, shows significant potential for improving fetal health assessment.
• The model provides a robust tool for early risk detection in obstetrics.
• Explainable AI features enhance trust and facilitate clinical adoption of AI-driven fetal health diagnostics.