Generalization of AI-Based Gestational Age Assessment Using Blind Sweep Ultrasonography

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Obstetrics

Background:

• Artificial intelligence (AI) systems are being developed for estimating gestational age (GA) using blind sweep ultrasonography.
• Evaluating the generalizability of these AI systems to new clinical settings is crucial.

Purpose of the Study:

• To investigate the generalizability of a previously trained AI system for GA estimation using blind sweep ultrasonography in new clinical environments.
• To assess the AI model's performance against the clinical standard for GA determination.

Main Methods:

• A prospective diagnostic study involving 2043 participants across two urban medical institutions (Chicago, USA and Nairobi, Kenya).
• A lightweight adaptation of the AI model was performed using data from one setting.
• Data collection utilized a mobile application with a protocolized blind sweep procedure by novice operators.

Main Results:

• The adapted AI model achieved a mean absolute error (MAE) of 4.1 days in Chicago and 4.3 days in Nairobi.
• Overall, the AI model demonstrated noninferiority to clinical standard GA estimates (MAE 4.2 days vs 4.5 days).
• The AI model showed generalization to new clinical settings and a different ultrasonography probe.

Conclusions:

• AI-based gestational age estimation using blind sweep ultrasonography is generalizable to new clinical settings.
• This technology has the potential to expand access to obstetric care, especially in low-resource environments.
• AI-powered ultrasonography offers a scalable solution for cross-setting deployment.