Inteligencia artificial, Machine Learning y Deep Learning en la medicina Fetal

Abstract

Traditional fetal medicine faces significant challenges due to observer variability and the operational burden of ultrasound diagnosis, prompting the search for advanced technological solutions. The aim of this study is to evaluate the scientific evidence on the effectiveness of artificial intelligence (AI) in prenatal diagnosis, risk prediction, and the identification of ethical and legal challenges. To this end, a systematic review was conducted under the PRISMA protocol, selecting 50 original articles from PubMed and SciELO after screening 12,000 initial records. The most relevant results show that deep learning architectures, such as convolutional neural networks, achieve up to 98.5% accuracy in classifying anatomical planes and surpass human performance in detecting cardiac and central nervous system anomalies. Furthermore, it is evident that AI reduces scan time by 42%, optimizing clinical efficiency. It is concluded that while AI transforms the paradigm towards proactive prevention and democratizes access to high-quality diagnoses, its implementation requires robust regulatory frameworks that mitigate algorithmic bias and define legal responsibilities, reaffirming its role as a complementary tool to human medical judgment.