Applications of artificial intelligence in fetal MRI: a systematic review

Abstract Background Magnetic resonance imaging (MRI) plays a pivotal role in obstetric care, offering high-resolution visualization for assessing fetal abnormalities, especially when ultrasound has limitations. Despite its advantages, MRI faces challenges such as motion artifacts and extended imaging durations, which can hinder diagnostic accuracy. The integration of artificial intelligence (AI) has demonstrated significant potential in overcoming these challenges by enhancing image quality, enabling precise segmentation, and automating diagnostic workflows. Continued evaluation and rigorous validation are essential to optimize AI’s clinical utility, ensuring its safe and effective application in prenatal care. Methods In November 2024, data were collected from three electronic databases—PubMed, Web of Science, and ScienceDirect—to explore AI applications in fetal MRI. Result Out of 1587 articles, 95 studies were included in this review. Key focuses include segmentation (37%), motion correction (22%), and disease prediction (18%). The fetal brain was the main anatomical target (53.6%). AI models like convolutional neural networks (CNNs) and U-Nets excel in segmentation and enhancement, with accuracy often over 90%. However, only 29.9% of studies provided data or code, and 69.8% included interpretability tools, highlighting reproducibility and clinical applicability challenges. Conclusion AI is transforming fetal MRI by improving segmentation, motion correction, and anomaly detection, especially through CNN models. Increased transparency, data access, and interpretability are crucial for clinical trust and integration into prenatal care, enhancing diagnostic precision and patient care.