Ultrasound imaging is the prominent analyzing method during pregnancy that used to measure some particular biometric parameters. It can be used for prenatal diagnosis and mainly estimating the gestational age. Fetal health circumference (HC) is one of the most integral part, that determines the fetus growth and health. This paper proposed a task for an automated segmentation that explores conventional neural network in depth. And evaluate HC ellipse by minimizing the cost function consisting of segment dice score and MSE of ellipticity parameters. Moreover, manual measurements are usually depended on the operator and it will take more time to evaluate the measurements. In fact, there are many researches on automated methods, but still there is a need for development in the system, in order to accuracy and reliability. This paper focus on the development of deep-learning method to evaluate the growth of fetal. The term fetal head biometry also deeply discussed in this paper. The proposed method trained in 110 labeled data set and examined 80 ultrasound images. The success rate for HC and BPD is 94.31% and in plane acceptance check achieved 89.14% of accuracy.

CNN, biometry, HC, nervous system, ultrasound image.