查看文章

Fetus Head Circumference is an important biometric parameter for detecting growth abnormalities and estimating gestational age. For prenatal screening, 2D ultrasound is preferred. Because of its disadvantages like missing boundaries and low signal to noise ratio, measurement of head circumference is time consuming and challenging. So, precise automatic segmentation is needed to help sonographers in efficient head circumference measurement. We propose an encoder decoder network having dense blocks that strengthen feature propagation and channel attention in bottleneck to enhance the important features. The network performs better than UNet. It has lesser inference time, takes 7 times lesser training time, 5 times lesser parameters and storage size than UNet. It can be used in real time because of the average inference time of 0.0037 seconds per batch of 4. The proposed network achieved an average …