AI-Enabled, Ultrasound-Guided Device Could Save Lives on the Battlefield

A team of researchers in the Massachusetts General Hospital Department of Radiology and at MIT Lincoln Laboratory have developed an ultrasound-guided handheld robotic device that allows non-specialists—including combat medics in battlefield settings—to access deep arteries and veins for life-saving applications. Dubbed AI-GUIDE, the device takes advantage of advances in artificial intelligence (AI) to enable semi-automated placement of large-bore catheters in patients in critical need of fluid replacement or other interventions. The researchers reported the device in December in the journal Biosensors.

Subscribe to the latest updates from Radiology Advances in Motion

Error: Please enter a valid email address.

Email Address Submit

In the Q&A below, Theodore Pierce, MD, an abdominal radiologist at Mass General, a faculty member in the Center for Ultrasound Research & Translation (CURT), and one of the inventors of AI-GUIDE, discusses the device itself, the applications that could benefit from its use, and the possibility of it bringing about a paradigm shift in ultrasound-guided interventions.

Q. What motivated your development of AI-GUIDE?

Pierce: The AI-GUIDE device was developed to meet the critical unmet need for point-of-care large-vessel cannulation by combat medics on the battlefield. Non-compressible hemorrhage is the leading cause of preventable battlefield death. Effective management requires the placement of large-bore vascular catheters into central vessels such as the internal jugular vein, subclavian vein, and femoral vein or major arteries. These catheters allow reliable delivery of large volumes of intravenous fluid resuscitation, administration of vasoactive medications, and the deployment of novel life support devices such as REBOA (Resuscitative Endovascular Balloon Occlusion) and portable ECMO (Extracorporeal Membrane Oxygenation).

Combat medics currently are not trained to perform large-vessel cannulation, which requires ultrasound interpretation capabilities and substantial experience with image-guided needle placement. Such procedures are typically performed by experienced physicians and specially trained advanced practitioners within a hospital setting. We sought to enable this capability for frontline medics by leveraging artificial intelligence and surgical robotics.

Q. How does the device help solve this problem?

Pierce: AI-GUIDE is designed to perform the two key components of vascular cannulation that are difficult for the average person: ultrasound image interpretation and precise image-guided needle placement. It accomplishes these tasks by integrating artificial intelligence ultrasound image analysis software, a surgical robotic needle delivery system, and the operator.

The AI software analyzes ultrasound images in real time in order to identify and localize blood vessels within the image. A simplified targeting display, with just a moving dot and target, translates the AI localization information to guide the operator to optimally position the device for attempted cannulation. In addition to the core targeting processes, several software logics have been implemented to improve the usability, accuracy and safety of the device. At this point, the surgical robotic system utilizes the AI targeting coordinates to plot an injection path. When ready, the operator depresses a button and the needle is inserted precisely within the vessel. The difficult part of vessel cannulation, 3D localization of the needle, is finished and all that remains is a one-dimensional problem of wire/catheter exchanges in order to place the desired final catheter. Ultimately, various application-specific needle cartridges will be developed to house an array of catheter types; this will allow single-step catheter insertion, remove the need for catheter exchanges, and allow delivery of a catheter specifically tailored for a given indication.

Q. What is most novel about AI-GUIDE?

Pierce: The novelty and functionality of the device are based on four key factors: AI algorithms, robotic integration, human dexterity and the broad potential of adaptation as a platform technology. Artificial intelligence is a rapidly growing field. We leverage recent advances in a GPU-enabled portable tablet's processing power, data annotation techniques, and real-time machine learning algorithms to develop highly dependable and fast algorithms to identify and segment large vessels using ultrasound images. The high contrast between blood vessels and background tissue, relative conservation of imaging appearance across subjects, and anatomic consistency facilitate AI functionality.

Direct integration of a robotic subsystem allows needle insertion to be performed precisely and consistently without requiring operator training or expertise. While typically, needle placement requires visualization of the needle throughout the procedure, our ability to precisely understand needle tip position based on external geometric factors provided by the surgical robotic system allows us to insert the needle from any angle, whether or not it is visualized throughout its injection path.

Other investigators have developed automated robotic systems for venipuncture, some of which take the human completely out of the procedure. In these cases, a large, complex robotic arm, or similar apparatus, is required to perform gross and fine positioning of the device. Our device is a critical departure from this mentality as AI-GUIDE works synergistically with the operator rather than seeking to replace them. We leverage the fact that most people possess the dexterity to hold the device and move it over a body surface, which obviates the need for overly complex robotic systems, which are costly and large, and preclude mobility. This allows our device to be substantially smaller and more mobile, thus facilitating its application on the battlefield, within emergency medical service systems, or for other point-of-care applications. This enables device deployment among frontline operators, those who would most benefit from AI-GUIDE. Lastly, the device is designed such that it can be rapidly adopted for adjacent applications through the utilization of application-specific loadable and disposable cartridges. This flexibility will maximize adoption and adaptation among a wide variety of indications.

Q. What was your role in the development of AI-GUIDE?

Pierce: What makes this device truly special is the unique combination of multiple complementary skillsets and disciplines leveraged for its creation. For my direct role in the project as a co-inventor and co-designer of the AI-GUIDE device, I provided clinical guidance to inform algorithm and hardware design features in order to mimic established clinical practices. Additionally, I helped develop standardized image acquisition algorithms in order to build an AI-ready ultrasound database for algorithm development while also designing device testing procedures on phantoms and a porcine model. I then assisted with human factors modifications of the device to improve the user experience. However, my greatest role was to function as a part of a team of clinical, engineering and computer science experts who, together, made this project successful. As a group, we look forward to translating this device into clinical care, expanding the indications for its utilization, and using the device to foster new innovative collaborations to improve care delivery.

Q. What other applications could benefit from the use of the device?

Pierce: While the device has been tailored to facilitate femoral vessel cannulation, minor software or cartridge modifications would permit accessing any blood vessel that can be visualized by ultrasound: for example, the internal jugular vein or upper extremity veins. The hardware within the cartridge can be optimized for the specific application by matching needle size or length to the specific vessel of interest. Likewise, the software can be tailored to perform complete vessel segmentation and targeting for novice users or permit manual override for experienced operators in selected settings.

Our future device is not limited to vascular access. Other candidate applications include percutaneous cricothyrotomy, pneumothorax decompression, body cavity fluid drainage, image-guided minimally invasive surgery, and image-guided targeted cancer therapy. While the comprehensive list of individual procedures is too long to list, conceptually, any procedure that involves ultrasound localization and targeting could be performed, facilitated, and optimized by AI-GUIDE and tailored based on operator skill and experience level.

Q. Do you have any thoughts about the broader significance of the device?

Pierce: Our group is incredibly excited by the prospect that this device may facilitate a paradigm change in ultrasound-guided intervention, which may open up new avenues of care delivery. With a patent pending and recent awards including a 2021 R&D 100 award and the 2021 MIT Lincoln Laboratory Best Invention award, we are pleased that others are equally excited. The device is elegant in its operational simplicity—identify a target, localize the target, hit the target—but the complexity "under the hood" reflects the hard work and dedication from a world-class multidisciplinary team with expertise in diagnostic imaging, minimally invasive procedures, computer science, artificial intelligence, mechanical engineering, phantom fabrication, veterinary expertise, and program management.

Q. What is the next step in the development of AI-GUIDE?

Pierce: The key next step in development is to refine the design according to strict FDA construction requirements to enable translation to first-in-human studies. Ensuring highly reliable component and subsystem performance, guaranteeing sterility, developing operator training modules, optimizing device-operator interface, and developing a human studies research plan with associated research subject safeguards will be core components of future work.

Learn more about the Center for Ultrasound Research & Translation (CURT)

Learn more about research in the Department of Radiology