Localizing Scan Targets from Human Pose for Autonomous Lung Ultrasound Imaging

Ultrasound is progressing toward becoming an affordable and versatile solution to medical imaging. With the advent of COVID-19 global pandemic, there is a need to fully automate ultrasound imaging as it requires trained operators in close proximity to patients for a long period of time, therefore increasing risk of infection. In this work, we investigate the important yet seldom-studied problem of ultrasound scan target localization, under the setting of lung ultrasound imaging. Existing works either lack human verification or generalization capability, by conducting experiments on phantom objects or relying domain-specific prior of the specific scan target. To address these issues, we develop a purely vision-based and data-driven method inspired by research in human pose estimation. We test the proposed method on 30 human subjects, and attain an accuracy level of 16.00±9.79mm for probe positioning and 4.44±3.75^∘ for probe orientation, with a success rate above 80% under an error threshold of 25 mm for all scan targets. Moreover, our approach can serve as a general solution to other types of ultrasound modalities. The code for implementation has been released.