37 36 HKU ANNUAL REPORT 2025 TECH TRANSFER AND KNOWLEDGE EXCHANGE IMPROVING ORAL HEALTH IN THE COMMUNITY The Faculty of Dentistry is harnessing AI in multiple projects to help vulnerable groups achieve better oral health. One project is the world’s first AI system to predict dental caries in the individual teeth of young children with an accuracy rate of more than 90%. The new system, developed by Professor Shi Huang and his team with the Chinese Academy of Sciences and other partners in the Chinese Mainland, paves the way for precision treatment of the most common childhood disease in the world. Another project is the world’s first AI-powered clinic for early oral cancer detection using the Faculty-developed tool OralCancerPredict, which has a 94% accuracy rate. The clinic, launched in April 2025 and led by Professor Richard Su, is a joint venture of the Faculty and Prince Philip Dental Hospital. A Faculty research team also developed an AI-assisted tool to analyse teeth and gum health, called GUM AI, that has been deployed to help the elderly in underserved communities, in a project involving community and industry partners in Hong Kong. GUM AI takes only a few seconds to complete its analysis and has an accuracy rate of over 90%. The Faculty has also been designated a Global Hub for Future Dentistry to drive dental innovation and entrepreneurship, in close collaboration with HKU’s TEC and the Hong Kong Science and Technology Parks Corporation, for the region’s unique theme-based dental-specific incubation programme. The research team, led by Professor Shi Huang (left), developed an AI system which achieved a 93% accuracy rate in predicting caries that would develop two months later. GETTING A FOOTHOLD A student-initiated start-up that uses AI and health technology to promote children’s foot health, won top awards at several major competitions in 2024–25. Digitoe Limited was founded by Raissa Yeung, who graduated in 2025 with a Bachelor of Social Sciences. She and her team created iWalk, an AI-powered smart insole that collects real-time walking data and gait analysis and integrates it with software to create personalised exercise recommendations. The information enables health professionals to detect and address foot and posture issues early on and facilitates the creation of custom-made insoles tailored to individual needs. The technology is advancing through the research and development phase, with formal clinical trials planned in close partnership with Queen Mary Hospital, Duchess of Kent Children’s Hospital and HKUMed’s Department of Orthopaedics and Traumatology. Among its honours, it was named Champion of Samsung Solve for Tomorrow 2024–25, Winner of the Hong Kong Trade Development Council’s Start-up Express 2025 and a Top 30 Start-up in Alibaba’s JUMPSTARTER 2025, beating out hundreds of entries in each event. iWalk developed by the team led by Raissa Yeung (first from left) offers a transformative solution to enhance children’s foot health, setting a new benchmark for AI health-tech innovation. Digitoe Limited was named Champion in the Tertiary Category of Samsung Solve for Tomorrow 2024–25. Scholars and students have launched inventions and start-ups that build on the new capabilities offered by artificial intelligence. AI-Powered Innovations FIRST AI MODEL FOR STAGING THYROID CANCER A groundbreaking model that uses four advanced large language models to streamline thyroid cancer management has been developed by a team from HKUMed. Thyroid cancer is a common malignancy, but effective treatment depends on determining the cancer stage and risk category. This determination has traditionally been done manually and is time-consuming – meaning the patient must wait longer for treatment. The new model was developed using the pathology reports, operation records and clinical notes from pseudo thyroid cancer cases or those from open-source datasets, which were fed into four offline large language models (LLMs) for privacy preservation. The output from the four LLMs was then assembled by the model. The impressive testing results found up to 100% accuracy in risk stratification and nearly 98% accuracy in staging. Importantly, the model reduces the diagnostic preparation time by nearly 50%. The model can also be readily integrated into various healthcare and research settings – public, private, local and international. The HKUMed team was led by Professor Joseph Wu Tsz-kei, Sir Robert Kotewall Professor in Public Health and Managing Director and Lead Scientist of InnoHK Laboratory of Data Discovery for Health (D24H), Dr Matrix Fung Man-him of the Department of Surgery, School of Clinical Medicine, and Dr Carlos Wong King-ho, Senior Research Director of InnoHK D24H. An HKUMed research team developed world’s first AI model for thyroid cancer diagnosis, with over 90% accuracy and reduced consultation preparation time. A research team from the Faculty of Dentistry has developed GUM AI, an innovative solution combining mobile health and artificial intelligence technology to analyse teeth and gum health.
RkJQdWJsaXNoZXIy ODI4MTQ=