Image-guided robotics for autonomous venipuncture

Performed over 1.5 billion times each year, venipuncture is the most ubiquitous clinical intervention and leading cause of medical injury in the United States. This dissertation describes the development of a medical device aimed at minimizing venipuncture-related adverse events by drawing blood and placing peripheral vascular catheters in an automated manner. The device may be particularly suited to challenging populations, including pediatric, geriatric, chronically-ill, and obese patients, for whom the rates of success depend heavily on practitioner skill. The device combines 3D near-infrared and ultrasound imaging, computer vision software, and a miniaturized dexterous robot that inserts the needle based on real-time image guidance. Methods to robustly segment, localize, and track the pose of the vessels are presented, and robotic prototypes capable of autonomously aligning and servoing the needle under real-time image guidance are introduced. In imaging studies on human volunteers spanning a broad range of demographic characteristics, the device was able to detect a majority of forearm veins that remained invisible to clinical visual assessment. In cannulation studies on in vitro and in vivo models, the device demonstrated significant improvements in first-stick accuracy and completion time compared to unassisted and image-guided manual venipuncture techniques. The results of these studies may be used to inform future clinical trials evaluating device feasibility in normal and difficult patient groups. Once translated, the device may be targeted to high-volume laboratory testing facilities and hospitals, and the underlying technologies may be extended to other diagnostic and interventional procedures, including automated point-of-care blood analysis, in-clinic and at-home hemodialysis venipuncture, peripherally-inserted central catheter placement, and emergency or military use.