Automated Detection of Common Maternal Fetal Ultrasound Planes Using Deep Feature Fusion

Abstract

Ultrasound is the primary imaging modality used to assess the development and well-being of the fetus during pregnancy. Identifying the right anatomical structure plays an important role to monitor the fetus development. However, identification of the right anatomical structure is a difficult and time-consuming process even for the skilled sonographer. Therefore, a deep learning-based automated detection system of common maternal fetal ultrasound planes using deep feature fusion is proposed. The deep attributes extracted from pretrained ResNet-50 and VGG-19-GAP are fused. These fused deep feature descriptors are given to the multiclass support vector machine to classify the fetal ultrasound planes into six classes such as the abdomen, brain, femur, thorax, cervix, and other planes. Experimental outcomes indicate that the developed multiclass categorization of fetal ultrasound planes using deep feature fusion outperforms existing state-of-the-art approaches in terms of accuracy.