Related Experiment Video
Updated: Aug 7, 2025
Deep Neural Networks for Image-Based Dietary Assessment
Published on: March 13, 2021
Human-guided deep learning with ante-hoc explainability by convolutional network from non-image data for pregnancy
Herdiantri Sufriyana1, Yu-Wei Wu2, Emily Chia-Yu Su3
1Graduate Institute of Biomedical Informatics, College of Medical Science and Technology, Taipei Medical University, 250 Wu-Xing Street, Taipei 11031, Taiwan; Department of Medical Physiology, Faculty of Medicine, Universitas Nahdlatul Ulama Surabaya, 57 Raya Jemursari Road, Surabaya 60237, Indonesia.
This study introduces a novel human-guided deep learning approach for predicting prelabor rupture of membranes (PROM) and delivery time. The model offers ante-hoc explainability, providing actionable insights for preventive medicine.
Area of Science:
- Medical informatics
- Artificial intelligence in healthcare
- Deep learning for predictive modeling
Background:
- Deep learning excels in medical imaging diagnostics but often lacks explainability.
- Current methods predominantly use post-hoc explanations, not integrated into model design.
- Explainable AI (XAI) is crucial for regulatory approval and clinical trust.
Purpose of the Study:
- To develop a prognostic prediction model for prelabor rupture of membranes (PROM) using ante-hoc explainable deep learning.
- To create an estimator for time of delivery from non-image data.
- To demonstrate a human-guided convolutional neural network approach for medical prognostication.
Main Methods:
- Constructed association diagrams from literature and EHR data to guide model development.
- Transformed non-image clinical data into image-like representations using predictor similarities.
- Employed convolutional neural networks (CNNs) with architecture inferred from data similarities for prognostication.
Main Results:
- Achieved AUCs of 0.73 (internal) and 0.70 (external) for PROM prediction in a nationwide database (n=883, 376).
- The developed model outperformed existing methods identified in systematic reviews.
- Model explainability was confirmed through knowledge-based diagrams and representation analysis.
Conclusions:
- The human-guided deep learning model provides accurate prognostication for PROM.
- Ante-hoc explainability enables actionable insights for preventive medicine.
- This approach advances the application of AI in clinical decision-making.