In Focus - Issue 35 (Spring 2023)

06 IN FOCUS In the News The “AI checker”, based on computer vision, helps to ensure all guidewires (circled, right) have been removed from a patient’s body. AI technology boosts patient safety An AI-based guidewire recognition and counting technology, developed by Prof. Gary CHAN Shueng-Han, Computer Science and Engineering, and his research team, is helping to improve patient safety by reducing the risk of wires being le in the body following a clinical procedure. The AI so ware, which rapidly and accurately recognizes all used guidewires removed from a patient, has been deployed at Tseung Kwan O Hospital in cases from November up to February , without missing a wire. Prior to the technology, medical workers had to manually count multiple times whether all guidewires had been taken out. A guidewire is a thin wire used to guide placement of a catheter into a central vein to administer fluids or medication, or monitor blood flow stability. It is used in procedures employing the Seldinger technique, where a central venous catheter is inserted to give access to blood vessels and hollow organs such as the stomach and gallbladder. The technology developed by Prof. Chan’s team provides an “AI checker” using computer vision. Medical sta rst con rm the number of guidewires as indicated by the system. They take and upload a photo of the guidewire(s), and perhaps other medical instruments, via a smartphone or tablet. Through object recognition and data augmentation techniques, the system can then detect guidewires and cross-check the manual count in real time. “Compared to humans, our AI checker does not get tired or distracted and serves as an impartial tool to double check and validate manual counting,” Prof. Chan said, noting the system maintains its high accuracy and e ciency over time. “Looking ahead, we would like to extend the technology from guidewires to other medical instruments to guard against retention inside a patient’s body.” Prof. Chan is also working on location sensing applications to e ciently search for missing dementia patients or other mentally incapacitated persons (MIPs). Turning innovation into industry impact Five HKUST advanced technologies are to be developed for application in industry following the signing of a licensing agreement with Guangdong Bright Dream Robotics Limited (BDR), which specializes in automation machinery R&D and manufacturing. The technologies mark the rst outputs to emerge from the HKUST-BDR Joint Research Institute (HBJRI), a University-industry platform set up to promote innovation and nurture talents. Established in , the joint institute is working on around projects related to robotics, AI, smart cities, materials, and big data. Most stem from real-life challenges facing industry, with faculty proposing solutions and Bright Dream Robotics testing the ideas. The innovations licensed comprise: A new inorganic coating that could help keep buildings cool in hot and humid climates by lowering the temperature by °C during day time and 6°C at night. A parallel compression framework containing three di erent compression algorithms to t di erent environments and point cloud data. GPU-accelerated algorithms can be used in large-scale D map compression or in real-time autonomous driving systems. Wear-resistant, anti-corrosion, self-cleaning, superhydrophobic surfaces for building robots’ metal components. Design and forming technology of large composites for building formwork and light-weight robots.