Artificial Intelligence-Augmented Pediatric Lung POCUS: A Pilot Study of Novice Learners

Area of Science:

• Pediatric Emergency Medicine
• Medical Imaging
• Artificial Intelligence in Healthcare

Background:

• Respiratory symptoms are common pediatric emergency department chief complaints.
• Point-of-care ultrasound (POCUS) is user-dependent and challenging for novice users.
• Chest X-ray is a conventional but less performant imaging modality.

Purpose of the Study:

• To introduce and assess the accuracy of AI-enhanced pleural sweep for lung imaging.
• To evaluate the diagnostic performance of novice learners using AI-assisted ultrasound for pneumonia identification.
• To compare AI-augmented lung ultrasound with expert interpretations.

Main Methods:

• Novice learners (NLs) received training on traditional lung POCUS and AI software.
• Previously healthy pediatric patients (0-17 years) with cardiopulmonary complaints were enrolled.
• Expert POCUS interpretations served as the criterion standard; NLs and experts were blinded.

Main Results:

• AI-augmented lung ultrasound achieved 66.7% sensitivity and 96.5% specificity for pneumonia detection by NLs.
• Overall accuracy for NLs using the AI tool was 93.7%.
• High interrater reliability (kappa=0.8) was observed between expert sonographers.

Conclusions:

• AI-augmented lung ultrasound demonstrates potential for improving diagnostic accuracy in pneumonia.
• The AI-assisted approach may enhance diagnostic efficiency for novice users in pediatric settings.
• This novel AI application shows promise for pediatric emergency medicine diagnostics.