Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AGE-US: automated gestational age estimation based on fetal ultrasound images

César Díaz-Parga, Marta Nuñez-Garcia, Maria J. Carreira, Gabriel Bernardino, Nicolás Vila-Blanco

TL;DR

GA estimation from fetal ultrasound is essential yet challenging when LMP is unavailable and manual measurements vary across operators. AGE-US introduces a lightweight, interpretable pipeline combining a shared-encoder head/abdomen segmentation network, a distance-map-based femur localization module, and a Hadlock-based GA calculation, leveraging transfer learning to cope with limited data. The approach achieves performance comparable to state-of-the-art methods while reducing computational complexity and enabling use in resource-constrained settings, with GA errors within inter-operator variability (MAPE around 2-3%). By mirroring clinical practice and using weak annotations for the femur, the method offers a practical, data-efficient route toward automated fetal biometry and GA estimation. The work highlights potential for broader deployment in low-resource environments, pending further validation and plane-selection automation.

Abstract

Being born small carries significant health risks, including increased neonatal mortality and a higher likelihood of future cardiac diseases. Accurate estimation of gestational age is critical for monitoring fetal growth, but traditional methods, such as estimation based on the last menstrual period, are in some situations difficult to obtain. While ultrasound-based approaches offer greater reliability, they rely on manual measurements that introduce variability. This study presents an interpretable deep learning-based method for automated gestational age calculation, leveraging a novel segmentation architecture and distance maps to overcome dataset limitations and the scarcity of segmentation masks. Our approach achieves performance comparable to state-of-the-art models while reducing complexity, making it particularly suitable for resource-constrained settings and with limited annotated data. Furthermore, our results demonstrate that the use of distance maps is particularly suitable for estimating femur endpoints.

AGE-US: automated gestational age estimation based on fetal ultrasound images

TL;DR

GA estimation from fetal ultrasound is essential yet challenging when LMP is unavailable and manual measurements vary across operators. AGE-US introduces a lightweight, interpretable pipeline combining a shared-encoder head/abdomen segmentation network, a distance-map-based femur localization module, and a Hadlock-based GA calculation, leveraging transfer learning to cope with limited data. The approach achieves performance comparable to state-of-the-art methods while reducing computational complexity and enabling use in resource-constrained settings, with GA errors within inter-operator variability (MAPE around 2-3%). By mirroring clinical practice and using weak annotations for the femur, the method offers a practical, data-efficient route toward automated fetal biometry and GA estimation. The work highlights potential for broader deployment in low-resource environments, pending further validation and plane-selection automation.

Abstract

Being born small carries significant health risks, including increased neonatal mortality and a higher likelihood of future cardiac diseases. Accurate estimation of gestational age is critical for monitoring fetal growth, but traditional methods, such as estimation based on the last menstrual period, are in some situations difficult to obtain. While ultrasound-based approaches offer greater reliability, they rely on manual measurements that introduce variability. This study presents an interpretable deep learning-based method for automated gestational age calculation, leveraging a novel segmentation architecture and distance maps to overcome dataset limitations and the scarcity of segmentation masks. Our approach achieves performance comparable to state-of-the-art models while reducing complexity, making it particularly suitable for resource-constrained settings and with limited annotated data. Furthermore, our results demonstrate that the use of distance maps is particularly suitable for estimating femur endpoints.

Paper Structure

This paper contains 9 sections, 1 equation, 3 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. Methodology
  4. Head and abdomen segmentation
  5. Femur localisation
  6. Gestational Age Estimation
  7. Experiments and Results
  8. Discussion
  9. Conclusions

Figures (3)

  • Figure 1: AGE-US workflow. First, abdomen and head masks are extracted using a single-encoder, dual-decoder U-Net, and a distance map of the femur endpoints is obtained using a conventional U-Net. Second, an ellipse is fitted to both the head and abdomen masks to calculate the corresponding circumferences (HC, AC), and the biparietal diameter (BPD), which is approximated by the minor axis of the head ellipse. In addition, the estimated femur endpoint location is extracted from the distance map, and its length is calculated. Thirdly, the 4 measurements are used in Hadlock's equation to compute the gestational age.
  • Figure 2: Representation of the method used to locate the femur endpoints (red crosses). Distance values are visualised using a colourmap for clarity, with blue representing the lowest (proximal) values and red representing the highest (distant) values.
  • Figure 3: Abdomen and head segmentation results.