In a move to improve the prompt detection and prevention of oral cancer, the Faculty of Dentistry has launched an AI clinic focussing on patients with potentially malignant oral disorders. Early Warning System We expect that the AI clinic will see a reduction in unnecessary surgery or overtreatment for otherwise low-risk oral leukoplakia and oral lichenoid disease patients who will be monitored regularly within the service. Easily accessible They chose internet deployment to make the AI tools – which are available free of charge – easily accessible to clinicians from any device, including smartphones, tablets or computers. The research started in 2019, when the researchers began to develop the AI-based models by collecting retrospective data of 716 patients with OPMD who had been treated at the Queen Mary Hospital and PPDH from 2003 to 2019. Tests in Hong Kong have shown that the models had an accuracy of more than 90 per cent. Further testing has been done afield – in the UK and Nigeria where good accuracy (over 80 per cent) among similar patients was also observed. “Since initial web tools were put up in March 2023, the platforms have been accessed more than 12,000 times from users in over 55 countries,” said Professor Adeoye. “Our recent external tests of OralCancerPredict in Hong Kong on 176 patients treated at PPDH in 2021 and 2022 found that it provided correct stratified risk in 87 per cent of patients, which was 11 per cent better than the current method of risk assessment as at last follow-up.” The researchers hope to improve treatment specifically for patients with oral leukoplakia (OL) and oral lichenoid diseases (OLD) by employing AI-based tools to provide personalised decision support to doctors. This AI-assisted workflow is crucial because treatment and monitoring strategies for OL and OLD depend on an initial cancer risk assessment by doctors, since not all patients will develop cancer in their lifetime (even without treatment). Tools like OralCancerPredict in the AI clinic will efficiently define oral cancer risk and streamline treatment with close monitoring of at-risk patients to promote disease prevention and early diagnosis. “Also, we expect that the AI clinic Patients’ feedback In order to reassure patients who may be suspicious about AI and worried that this is ‘treatment by robot’, the team were careful during the model development stage to sample patients’ reactions on how they perceived their doctors if they employed the results of an AIbased model to guide their treatment decisions. “Many were not against it and highlighted that they would provide consent as long as the AI results were not the primary determinant of treatment and only supported doctors’ clinical judgments,” said Professor Adeoye. “We will continue to inform our patients that final decisions on treatment and follow-up strategies are made by the doctors pending confirmation from patients.” Other dental faculties have already reached out to learn more about how they could benefit from using AI tools. Professor Adeoye said: “We expect the uptake to increase further in the city when data from our clinic shows the impact of the AI tool on clinical decision-making and OPMD patient outcomes.” In the meantime, the team have also developed other accurate and validated AI tools such as DysPOLNet for the comprehensive management and monitoring of OPMD. will see a reduction in unnecessary surgery or overtreatment for otherwise low-risk OL/OLD patients who will be monitored regularly within the service,” he said. One of the outputs of OralCancerPredict is binary risk stratification. Used this way, the model can stratify patients into high- and low-risk groups according to the probability of cancer development. OL/OLD patients with high risk will be monitored in the AI clinic every three months (close monitoring), while patients with low risk will be monitored in the AI clinic every six months. Professor John Adeoye The AI-powered clinic was developed by the Division of Oral and Maxillofacial Surgery (OMFS) of HKU’s Faculty of Dentistry and the Prince Philip Dental Hospital (PPDH). It is a culmination of research efforts to improve risk assessment, treatment selection and monitoring for patients with oral potentially malignant disorders (OPMD). “The AI clinic was launched as a new medical service at the OMFS Clinics of PPDH in April this year. It was our response to the need for better treatment and monitoring of patients, since early detection is crucial to survival rate in oral cancer,” said Professor Richard Su, Chief and Clinical Professor of the OMFS, who leads the AI clinic with contributions from Professor Liwu Zheng, Professor Jane Jingya Pu and Dr Joanna Ko from the OMFS, and Professor John Adeoye, Assistant Professor in Digital and Precision Dentistry. The team felt that the current method for assessing cancer development in OPMD is outdated and unidimensional. Multidimensional factors have to be taken into consideration when assessing patients with OPMD, including demographics, risk habit history, clinical factors, and histological factors. “Until now, all of these factors were assessed separately and this was particularly challenging for busy general practitioners. Therefore, we developed and deployed the online tools powered by AI to improve the efficiency and clinical management of patients. One such AI tool termed OralCancerPredict integrates these factors to provide better risk assessment than methods previously applied in clinical settings,” said Professor Su. HKU Bulletin | Nov 2025 Knowledge Exchange 36 37
RkJQdWJsaXNoZXIy ODI4MTQ=