FetalMind: AI System for Fetal Ultrasound Report Generation and Diagnosis

A new medical AI system called FetalMind has been introduced for fetal ultrasound interpretation, capable of both report generation and diagnosis. The system addresses challenges in fetal ultrasound such as multi-view image reasoning, numerous diseases, and image diversity, which are not well handled by existing medical vision-language models adapted for adult imaging. FetalMind incorporates a Salient Epistemic Disentanglement (SED) mechanism that uses an expert-curated bipartite graph to decouple view-disease associations and guide preference selection via reinforcement learning, aligning the model's inference with obstetric practice. The system is trained on a curated dataset called FetalSigma, which is mentioned in the abstract. The research was announced on arXiv with the identifier 2510.12953v4, and the paper is available at https://arxiv.org/abs/2510.12953.