This study aims to develop advanced artificial intelligence (AI) models that predict neonatal risks and complications based on historical multimodal health data, including ultrasound and MRI scans. The objective is to empower clinicians and provide clear, compassionate support for families navigating complex prenatal diagnoses.

The FetalFirst study employs observational, retrospective analysis utilizing DenseNet121 neural networks. It analyzes de-identified retrospective data comprising ultrasound images, MRI scans, and clinical documentation from existing medical records. This research has received ethical approval from Wales Research Ethics Committee (REC ref: 25/WA/0168, IRAS ID: 358793). Outcomes from this study are expected to significantly enhance clinical intervention strategies, offering healthcare professionals robust tools for earlier detection and improved management of congenital anomalies and neonatal risks. Additionally, the insights gained will provide critical support to parents facing high-risk pregnancies, assisting them in making informed decisions.

Inclusion Criteria:

• Historical, de-identified neonatal records including ultrasound images, MRI scans, and clinical documentation available for analysis.

Exclusion Criteria:

• Cases with incomplete or missing critical data elements required for AI model analysis.

The population includes retrospective, anonymized neonatal records obtained from clinical datasets, focusing on neonates previously diagnosed or assessed for congenital anomalies and neonatal risks.

Individual participant data (IPD) will not be shared due to ethical and privacy considerations. All data is pseudonymised and governed strictly by the approved Data Sharing Agreement, which restricts external sharing to protect participant confidentiality and comply with UK GDPR.