Clinical Validation of an Artificial Intelligence System for De-termining Fetal Lie and Position in 3rd Trimester

Overview

This prospective clinical validation study aims to evaluate the diagnostic performance of an artificial intelligence (AI)-based navigational support system for determining fetal lie and presentation during third-trimester ultrasound examinations. Following routine clinical assessment by an expert clinician, three blind ultrasound sweeps will be obtained from each participant. The AI system will generate predictions of fetal lie and presentation from each sweep, which will be compared with the expert clinician's assessment as the reference standard. The study will also explore clinician perceptions of the system's usability and potential clinical value.

Accurate assessment of fetal lie and presentation during the third trimester is essential for obstetric management and delivery planning. Ultrasound is the gold standard for determining fetal orientation; however, the examination is highly operator-dependent and requires considerable experience to correctly identify fetal anatomy and orientation. Artificial intelligence (AI)-based navigational support systems have the potential to assist clinicians by providing real-time guidance during ultrasound examinations and improving the consistency of fetal orientation assessment. The purpose of this study is to prospectively validate an AI-based navigational support system for determining fetal lie and presentation during routine third-trimester ultrasound examinations. The system analyzes blind ultrasound sweeps acquired along the maternal midline and automatically predicts fetal lie and presentation. AI-generated predictions will be compared with expert clinician assessment, which serves as the reference standard. Eligible pregnant women attending third-trimester ultrasound examinations at the Fetal Medicine Department, Rigshospitalet, Slagelse - or Hillerød hospital, will be invited to participate. Following written informed consent, an expert clinician will first determine fetal lie and presentation as part of the routine ultrasound examination. The same clinician will subsequently acquire three blind ultrasound sweeps along the maternal midline. The AI navigational support system will analyze each sweep and generate predictions of fetal lie and presentation. All examinations will be completed within 5-10 minutes to minimize changes in fetal orientation between the clinical assessment and AI evaluation. The primary objective is to evaluate the diagnostic performance of the AI system by comparing AI-generated predictions with expert clinician assessment. Primary outcome measures include overall diagnostic accuracy and agreement, These findings will provide additional insight into the feasibility of implementing AI-assisted navigational support in routine obstetric ultrasound practice.