Related Experiment Video
Updated: Jan 3, 2026
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.3K
Postnatal gestational age estimation of newborns using Small Sample Deep Learning
Mercedes Torres Torres1, Michel Valstar1, Caroline Henry2
1School of Computer Science, University of Nottingham, United Kingdom of Great Britain and Northern Ireland.
Image and Vision Computing
|November 26, 2019
Summary
Accurate gestational age estimation for newborns is crucial for timely treatment. This study introduces an AI system using facial, foot, and ear images, achieving high accuracy and outperforming traditional methods, especially in remote areas.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Neonatal Care
Background:
- Accurate gestational age assessment is vital for neonatal care and determining prematurity.
- Ultrasound scans (USS) are accurate but costly and inaccessible in remote regions.
- The Ballard Score is a subjective clinical alternative with variable accuracy based on examiner experience.
Purpose of the Study:
- To develop an automated system for precise postnatal gestational age estimation.
- To overcome limitations of current methods, particularly in resource-limited settings.
- To leverage neonatal imaging and clinical data for improved accuracy.
Main Methods:
- A novel two-stage deep learning architecture was employed.
- Convolutional Neural Networks (CNNs) predicted broad gestational age classes from facial, foot, and ear images.
- Support Vector Regression (SVR) fused CNN outputs with infant weight for fine-grained predictions.
Main Results:
- The system achieved an expected error of 6 days, surpassing existing automated methods.
- It demonstrated three times greater accuracy than the Ballard Score.
- Incorporating images improved prediction accuracy by 33% compared to using weight alone.
Conclusions:
- The developed system offers a viable, accurate, and accessible alternative for postnatal gestational age estimation.
- This AI-driven approach is particularly beneficial for areas lacking advanced medical imaging facilities.
- The method shows significant potential to improve neonatal care outcomes globally.
