A comprehensive review of artificial intelligence - based algorithm towards fetal facial anomalies detection (2013–2024)

Abstract

This review explores the growing need for AI-based algorithms in diagnosing fetal facial anomalies, which are often difficult to detect due to limitations in current imaging techniques like ultrasound and MRI. These challenges include low resolution, motion artifacts, and insufficient annotated data, which hinder early and accurate diagnosis. AI algorithms, particularly deep learning models like Convolutional Neural Networks (CNNs) and U-Net, offers significant potential to overcome these challenges by analyzing large datasets and improving image analysis. Early diagnosis of these anomalies is crucial for enabling timely interventions, personalized treatment plans, and better prenatal care. This study adopts a systematic review approach, to assess existing research on AI-based approaches for fetal facial anomaly detection. The review includes peer-reviewed studies from key biomedical databases like PubMed, IEEE Xplore, and ScienceDirect, focusing on the last 15 years. Studies that implemented AI techniques and manual techniques for detecting anomalies in prenatal images were considered. Among all models reviewed, CNNs and U-Net architectures were found to be the most effective. CNNs excel at classifying medical images, while U-Net is particularly powerful for image segmentation. These models have demonstrated high accuracy in identifying conditions such as cleft lip, palate, and micrognathia. The use of AI in clinical settings can greatly enhance the precision and efficiency of fetal anomaly detection, addressing current limitations in medical imaging. By integrating AI, particularly deep learning models, into clinical workflows, prenatal care can be transformed, allowing for earlier and more accurate diagnosis. This can lead to more personalized care, timely interventions, and ultimately improved health outcomes for affected individuals and their families.