Automated Segmentation and Trimester-Based Classification of Fetal Head Circumference in Ultrasound Images Using Deep Learning Techniques

Ultrasonography is a noninvasive imaging technique widely used to monitor fetal growth, anatomy, and development. Among fetal biometric parameters, the head circumference (HC) is a crucial indicator of brain development and gestational age estimation. This study presents an optimized deep learning framework for automatic segmentation and trimester-wise classification of fetal HC ultrasound images. The framework was evaluated on two datasets: the public HC-18 dataset and a newly curated HC24 dataset encompassing images from all three trimesters. Multiple deep learning architectures, including conventional convolutional neural network (CNN)-based and transfer learning models, were systematically compared for segmentation performance. The best-performing model was subsequently integrated into the classification stage. To enhance generalization and mitigate overfitting, data augmentation and model fine-tuning strategies were applied. Experimental results demonstrate that the proposed framework achieves robust and clinically relevant performance, showing significant improvements in segmentation and classification metrics, including Accuracy, Precision, Recall, and F1-score across both datasets. These findings confirm the efficacy of the framework for fetal HC assessment and provide a reliable tool for trimester-wise fetal evaluation in ultrasound imaging.