An Ultrasound Transducer Tracking System Enhanced by Artificial Intelligence: A Camera-Based Approach

Abstract

Acute respiratory infections (ARIs) are a leading global health issue, accounting for significant child mortality. In 2022, Peru reported a 282% surge in ARI cases, largely attributed to the third COVID-19 wave. Traditional ARI diagnostic methods, prevalent in urban regions, remain resource-intensive, causing diagnostic challenges in rural settings. Ultrasound (US) imaging, especially Volume Sweep Imaging (VSI), emerges as a potential alternative due to its affordability, portability, and high efficacy in detecting ARIs. This study introduced an AI-aided camera-based US transducer tracking system for VSI, aiming to make US imaging accessible to rural Peru, where a lack of trained personnel exists. The system captures, processes, and classifies US acquisitions using three Logitech C925E cameras focusing on the anatomical planes and YOLOv5 as the object detection algorithm. Experimentation showed capability of the system to measure the US probe speed with under 4% error and to accurately describe its orientation. Additionally, the system demonstrated an accuracy of 96% in distinguishing between correct and incorrect USs. This system promises to bridge the diagnostic gap in underserved regions, due to its effectiveness on measuring US probe speed and distinguishing correct US acquisitions.