CAD System Utilizing UNet and Hough Transform for Automated Measurement of Fetal Head Circumference and Age in 2D Ultrasound Images

Hamzah Jaber; Ahmed Abed Mohammed; Bo Zhang; Maidi Qiu; Mustafa M. Abd Zaid; Putra Sumari

doi:10.47852/bonviewAIA52025416

Authors

Hamzah Jaber General Directorate of Vocational Education, Vocational Education Department, Iraq
Ahmed Abed Mohammed College of Computer Science and Information Technology, University of Al-Qadisiyah, Iraq
Bo Zhang School of Computer Science, Universiti Sains Malaysia, Malaysia
Maidi Qiu Haojing College, Shaanxi University of Science and Technology, China
Mustafa M. Abd Zaid College of Technical Engineering, Islamic University, Iraq
Putra Sumari School of Computer Science, Universiti Sains Malaysia, Malaysia

DOI:

https://doi.org/10.47852/bonviewAIA52025416

Keywords:

fetal ultrasound, segmentation, head circumference, UNet, Hough transform

Abstract

Two-dimensional (2D) medical ultrasound is a widely used imaging modality for the anatomical and functional assessment of fetal development due to its low cost, availability, real-time capability, and the absence of radiation hazards. Head circumference (HC) is an essential biometric to measure fetal growth. However, the low signal-to-noise ratio in ultrasound imaging can make it difficult for clinicians to identify the fetal plane correctly. Additionally, manually measuring HC can be expensive, involving accurately placing three minor and major parameter points from the ultrasound machine. To address these issues, research has been conducted to develop an automated system for measuring HC. This study presents a computer-aided diagnosis (CAD) system for the automatic measurement of fetal HC and fetal age using hybrid feature extraction. Using Convolutional Neural Networks (CNNs), self-supervised learning (SSL), vision transformers (ViTs), UNet deep learning model for segmentation, and Hough transform to measure performance, this study achieved higher performance compared to previous studies with a Dice similarity coefficient (DSC) of 97.23 ± 2.78, an average distance factor (ADF) of 2.8 ± 2.93 mm, a Jaccard Index of 88.57 ± 3.79, and an accuracy of 97.2%. After that, we enhance UNet using an attention mechanism that achieved a Dice coefficient of 98.5 ± 2.5, an ADF of 2.4 ± 2.8 mm, and an accuracy of 98.1%. This system provides a more cost-effective and accurate measurement of HC, aiding clinicians in assessing fetal development.

Received: 14 February 2025 | Revised: 16 May 2025 | Accepted: 29 July 2025

Conflicts of Interest

The authors declare that they have no conflicts of interest to this work.

Data Availability Statement

The data that support the findings of this study are openly available in GitHub at: https://github.com/pranjalrai-iitd/Fetal-head-segmentation-and-circumference-measurement-from-ultrasound-images.

Author Contribution Statement

Hamzah Jaber: Conceptualization, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing. Ahmed Abed Mohammed: Methodology, Validation, Formal analysis, Writing – review & editing. Zhang Bo: Software. Qiu Maidi: Software. Mustafa M. Abd Zaid: Validation, Project administration. Putra Sumari: Visualization, Supervision, Project administration.