Robotic Acquisition and Enhancement of Ultrasound Images

Abstract

Ultrasound is one of the cheapest, safest, and most ubiquitously found medical imaging technology. However ultrasound imaging is currently underutilized in image-guided interventions due to limited image quality and the lack of automated methods to provide image-based guidance. This work describes methods to improve ultrasound image quality and enables new procedure guidance methods through robotic steering of ultrasound imaging catheters. The first two chapters present a robotic system that was designed and built to autonomously steer ultrasound-imaging catheters. This system enables autonomous tracking of instruments inside the heart, and can generate 4D (3D + time) ultrasound volumes from the 2D images acquired from the catheter. This work improves visualization of cardiac structures during interventions, potentially leading to a decrease in the use of X-ray fluoroscopy. Next two chapters present methods for improving ultrasound image quality. One method increases interpolation accuracy during ultrasound scanline conversion using Gaussian process regression. The second method extends the dynamic range of ultrasound images through high dynamic range imaging methods, similar to how smartphones and digital cameras can combine pictures taken at different exposures to improve the visibility of different structures. The final chapter describes a platform that was developed to characterize the dynamic response of human muscles and internal organs, such as the stomach. The data collected in these studies can help improve the fidelity of finite element models of the human body.