<p>This work is intended to provide a comprehensive analysis of deep learning applications in fetal surveillance using ultrasound images while highlighting the current trends in the research and challenges in this field. Our work includes a decade-long survey of studies ranging from 2015 to 2024. This includes explicitly studies focusing on deep learning techniques and excluding those based on traditional methodologies like machine learning approaches. Ninety-one studies were analyzed, covering different fetal surveillance categories such as the fetal head, face, heart, and abdomen. According to our findings, the fetal head is the most extensively studied category among others in fetal surveillance, likely due to the availability of the publicly accessible HC18 dataset. Throughout the studies, we analyzed that not many datasets for fetal surveillance are publicly accessible, making it a significant challenge in the field. We observed that many ultrasound-imaging modalities are used for fetal surveillance, including 2D, 3D, and even 4D; however, 2D ultrasound images are utilized more frequently due to their accessibility and ease of interpretation. This survey outline the areas most investigated in fetal surveillance, such as the specific deep learning techniques employed, the availability of datasets in each category, the evaluation techniques used, and the specific tasks performed within each category. The key findings are highlighted and summarized in tables, which provide detailed information on study objectives, datasets, deep learning techniques, gestational age (GA), performed tasks, and achieved performance metrics. This review summarize recent advancements emphasizing limited dataset accessibility, clinical applicability, lack of explainability, and interpretability of predictive models. It also suggest future directions to enhance automatic fetal surveillance systems through deep learning.</p>