Akademik Veri Yönetim Sistemi
BIOMEDICAL SIGNAL PROCESSING AND CONTROL, cilt.119, sa.A, ss.1, 2026 (SCI-Expanded, Scopus)
Fetal head circumference (HC) measurement is an important parameter for the evaluation of fetal development during pregnancy. However, manual measurement methods have problems such as expert dependency and measurement inconsistency due to the characteristics of ultrasound images. In this study, we propose a hybrid deep learning model, assisted by transfer learning, which combines a ResNet-50-based DeepLabV3 network with a UNet3+ architecture to automatically and accurately detect fetal head circumference from ultrasound images. As a result of the analysis performed on the HC18 dataset, the model achieves 97.83% Dice similarity coefficient (DSC), 0.67 ± 2.62 mm difference (DF), 1.96 ± 1.86 mm absolute difference (AD) and 1.30 ± 0.79 mm Hausdorff distance (HD), showing that segmentation success is directly reflected in measurement accuracy and can compete with existing approaches in the literature. Notably, the model achieves this performance without requiring data augmentation or complex pre/post-processing steps and with a training time of approximately 20 min, demonstrating practical efficiency and reduced computational overhead. Ellipse fitting is performed solely on the segmentation output to obtain the final HC measurement. The proposed method can be used as a reliable and effective tool for fetal ultrasound image segmentation and head circumference measurement.