Prospective Testing of AI-Enhanced pOCUS for Pediatric Wrist and Elbow Fracture Detection

Introduction

Artificial intelligence (AI) is transforming medical imaging by enhancing diagnostic accuracy, streamlining workflows, and improving accessibility. Point-of-care ultrasound (pOCUS) has become an essential tool in emergency and clinical settings, offering real-time imaging without radiation exposure. However, its effectiveness is highly dependent on operator expertise, leading to variability in interpretation. AI-driven imaging solutions have the potential to address this challenge by standardizing ultrasound interpretation and augmenting clinical decision-making. This study evaluates the performance of an AI-enhanced pOCUS tool for detecting pediatric wrist and elbow fractures in a high-volume emergency setting. By assessing AI-assisted imaging against traditional radiographic modalities, we aim to determine its diagnostic accuracy, clinical integration, and potential impact on workflow efficiency.

Methods

We conducted a multi-center, prospective observational study evaluating AI-assisted pOCUS in pediatric patients presenting with suspected wrist or elbow fractures at the Stollery Emergency Room and affiliated clinics. Patients underwent standard clinical imaging, including X-rays, and when clinically indicated, CT or MRI scans. AI-assisted pOCUS was performed by trained healthcare professionals using an AI-powered interpretation tool. The AI model analyzed ultrasound images in real time and generated predictive assessments of fracture presence, severity, and anatomical location. These AI-generated diagnoses were then compared to radiologist-confirmed findings from X-rays and advanced imaging. The study evaluated the diagnostic performance and clinical integration of AI-assisted point-of-care ultrasound (pOCUS) in pediatric fracture detection. Sensitivity and specificity were assessed to determine AI accuracy compared to gold-standard imaging. Agreement between AI predictions and radiologist diagnoses was analyzed for consistency. Usability and integration in clinical settings were explored, considering ease of use, provider confidence, and barriers to implementation.

Results

Preliminary analysis suggests that AI-enhanced pOCUS demonstrates promising diagnostic performance, with early findings indicating high agreement with radiologist-confirmed diagnoses. The AI model efficiently identified key fracture characteristics, such as displacement and involvement of growth plates, contributing to faster and more standardized image interpretation. Additionally, workflow integration was assessed by measuring the time required for AI-assisted diagnoses compared to conventional imaging workflows. AI-assisted pOCUS showed potential in reducing time-to-diagnosis, particularly in resource-limited settings where access to radiologists may be delayed. Clinicians reported improved confidence in ultrasound-based diagnoses when AI support was available, particularly among less experienced operators. Further statistical analysis is ongoing to determine the sensitivity, specificity, and overall accuracy of AI-assisted pOCUS relative to X-rays and advanced imaging modalities.

Conclusion

AI-assisted pOCUS holds significant potential in enhancing pediatric fracture detection, streamlining diagnostic workflows, and reducing reliance on traditional imaging, particularly in emergency settings. By improving interpretation accuracy and accelerating clinical decision-making, AI integration could lead to faster treatment initiation, reduced patient wait times, and greater healthcare efficiency. If validated, this approach may support broader AI adoption in point-of-care diagnostics, with implications for training, resource allocation, and equitable access to imaging services. Future directions include expanding the dataset, refining AI algorithms for greater precision, and evaluating long-term clinical outcomes associated with AI-assisted ultrasound in pediatric care.