Hafiz Muhammad Danish ^1* Zobia Suhail ¹ Faiza Farooq ²

• ¹ University of the Punjab, Lahore, Pakistan
• ² University of Lahore, Lahore, Punjab, Pakistan

Monitoring the morphological features of the gestational sac (GS) and measuring the mean sac diameter (MSD) in early pregnancy serve to predict spontaneous miscarriage and approximate gestational age (GA). The manual process is laborious and relies significantly on the sonographer's skill. As a result, an automated pipeline was created to aid sonographers in precisely segmenting the GS and determining GA. The process began with preparing a novel dataset of 500 ultrasound (US) scans taken between 4 and 10 weeks of gestation. Four widely used fully convolutional neural networks: UNet, UNet++, DeepLabV3, and ResUNet were then modified by replacing their encoders with a pre-trained ResNet50. After training, the results were analyzed, identifying the ResUNet model as the optimal approach for segmenting the GS. Following this, novel biometry was introduced to assess GA automatically. Finally, the system's performance was compared with that of sonographers. Our experiments, which utilized 5-fold cross-validation, consistently demonstrated the superior performance of ResUNet with mean Intersection over Union (IoU), Dice, Recall, and Precision values of 0.946, 0.978, 0.987, and 0.958, respectively. The discrepancy between the sonographer's estimated GA and our biometry algorithm's estimations was measured at a Mean Absolute Error (MAE) of 0.07 weeks. These findings indicate that the proposed pipeline provides a more precise and reliable alternative to conventional manual measurements. Moreover, its applicability extends to segmenting and measuring other fetal components in future studies.