On the afternoon of May 15, 2025, the "International Forum on Medical Engineering Intersection 2025" was held in Room 210 of East Building No.5. This event was co-organized by the School of Computer Science and Technology at Huazhong University of Science and Technology(HUST), the University of New South Wales(UNSW), the University of Sydney(USYD), and the National University of Singapore(NUS). The forum focused on cutting-edge topics in medical engineering intersection, bringing together top scholars from China and Australia. It attracted participants from around the globe, including scholars, healthcare professionals, and technology experts, who engaged in in-depth discussions on key issues such as the application of artificial intelligence in healthcare, federated learning system design, distributed systems, and large model technologies. The discussions explored how AI technology is reshaping the global healthcare landscape and promoting interdisciplinary collaboration and innovation. The meeting was chaired by Associate Professor Feng Lu from HUST, with participation from Vice Dean Ruixuan Li and over 40 students.
During the opening ceremony, Vice Dean Ruixuan Li of the School of Computer Science and Technology at HUST delivered a welcoming speech, extending warm greetings and sincere thanks to the attending experts, scholars, and students. He reflected on the achievements of past collaborations while highlighting that the deep integration of artificial intelligence and healthcare is transforming traditional medical models. From privacy protection in federated learning to the clinical implementation of distributed systems, and from large model-enabled diagnostics to the deep mining of unstructured medical data, every breakthrough represents a convergence of computational science and medical expertise. He expressed hope for continued international exchanges and resource sharing to address key challenges in healthcare intelligence, promote the development of medical engineering methodologies, and inject new vitality into global healthcare.
In the lecture segment, Professor Louisa Jorm, Founding Director of the Health Big Data Research Center at UNSW, presented the topic "Beyond Algorithms—Returning Health AI to Clinical Reality." She noted that medical AI research has long been trapped in a "model performance race," excessively pursuing marginal improvements in metrics such as AUC while neglecting the real complexities of healthcare systems. Through case analyses, she illustrated that patient pathways from community clinics to emergency rooms and hospitalizations are nonlinear, with severe data fragmentation and often missing key health determinants like behavioral habits and social support. Her team’s "Heart AI" project, utilizing multi-center data integration and interpretable models, significantly improved the accuracy of readmission predictions for heart failure patients, reflecting a hopeful vision for transformative changes in healthcare intelligence through interdisciplinary collaboration.
Professor Albert Y. Zomaya, a Chief Professor at USYD and a member of the European Academy of Sciences, shared insights on "Solving Medical Challenges through Distributed Systems and Generative Artificial Intelligence." He focused on the integration of distributed systems and artificial intelligence, systematically outlining the evolution of distributed systems. In the generative AI section, he detailed the potential of cutting-edge technologies, such as GANs and VAEs, in medical data completion and augmentation. By providing practical case studies, he demonstrated how these technologies effectively fill gaps in medical records, thereby offering a more solid foundation for healthcare decision-making. In a highly sensitive medical industry concerning data privacy and security, federated learning enables multiple healthcare institutions to collaboratively train AI models without sharing raw data, thus protecting patient privacy while facilitating knowledge sharing. Although the prospects for healthcare intelligence development are promising, challenges remain, including enhancing model interpretability, addressing data bias, ensuring privacy protection, and tackling ethical issues related to fairness. Continuous attention and exploration are needed within the AI healthcare ecosystem.
Professor Bingsheng He, Vice Dean of the School of Computing at NUS, shared insights on the importance of data protection with his presentation titled "Federated Learning Systems: Effective and Efficient Machine Learning Systems for Data Silos." He explained the concept of "data silos" and highlighted the significance of federated learning for enabling collaborative training of machine learning models across different organizations while respecting privacy constraints. However, he noted that federated learning faces various challenges, such as unrealistic system assumptions, scalability, and efficiency issues. Furthermore, existing federated learning systems rarely consider two important characteristics of other federated systems: heterogeneity and autonomy. His research explores future opportunities and directions for data systems by comparing existing federated learning frameworks.
Dr. Oscar Perez Concha, a Researcher and Senior Lecturer at the Centre for Big Data Research in Health at UNSW, presented "Unlocking Unstructured Health Data: From Challenges to Breakthroughs—Part I." He began by emphasizing the significance of structuring clinical text and automatic clinical coding. He pointed out that clinical texts contain rich medical information, yet they often exist in unstructured formats, making them difficult to use for data analysis and decision support. The goal of automatic clinical coding is to transform these unstructured texts into standardized medical codes, achieving data structuring and standardization. He elaborated on his research team's methodologies in automatic clinical coding and text summarization, highlighting key outcomes such as improved model performance, effective text summarization, and enhanced interpretability, while reiterating the critical role of interdisciplinary collaboration in the success of medical AI.
Qing Ye, Deputy Director of the Big Data and Artificial Intelligence Office at Tongji Hospital, HUST, introduced a classification fusion model based on "Ideal Solution Similarity Ranking Technology" in his presentation titled "AI-Based Multidisciplinary Predictive Model for Invasive Shock Mortality: A Multicenter Retrospective Study." This model integrates seven machine learning models using TOPSIS, achieving an AUC of 0.733 in internal validation, with pediatric ICU and respiratory ICU AUCs of 0.808 and 0.662, respectively. External validation AUCs were 0.784 and 0.786, demonstrating high stability and accuracy across specialties and multicenter validations. This interpretable model provides clinicians with a reliable early warning tool for predicting the risk of septic shock mortality, aiding in early interventions to reduce mortality rates.
Dr. Wei Li, a Researcher at the Distributed and High-Performance Computing Centre at USYD, addressed the complex and highly individualized experience of pain in his presentation titled "Reconsidering Pain Management from a Computational Perspective." He noted that pain is a subjective experience influenced by various factors, including physical injury, emotions, stress, and fatigue. Current self-report tools commonly used in hospitals, while simple, have significant limitations, as they only capture pain at a specific moment and fail to reflect its dynamic changes, often influenced by subjective patient factors. The integration of computational technology offers new hope for pain management; by combining electronic health records and physiological signals collected from wearable devices, AI and machine learning can uncover patterns and changes that may be overlooked. The key to intelligent pain management lies not in the algorithms or computational systems themselves, but in ensuring that models have clinical significance and practicality at the point of care. Future research will focus on integrating structured records with continuous signals to achieve closed-loop control and develop models that are not only accurate but also interpretable, reliable, and deployable in clinical settings, ultimately enhancing patients' quality of life.
Finally, Professor Qinbin Li from HUST presented on "Distributed Systems in the Age of AI." He pointed out that traditional distributed systems primarily handle general computing tasks; however, with the rise of machine learning, particularly deep learning, distributed systems are increasingly supporting model training. As large-scale models become widely used, the focus of distributed systems will shift to supporting various intelligent applications based on these models. He showcased the tremendous potential of distributed systems in healthcare through case studies of medical applications based on distributed agent systems, providing valuable insights for the integration of distributed systems and AI, demonstrating cutting-edge technological advancements, and expanding new directions for future intelligent applications.
The successful hosting of this forum not only reflects the friendly cooperative relationships between HUST and international partners such as USYD, UNSW, and NUS but also establishes a regular, high-level international academic exchange platform. This initiative promotes the development of medical engineering methodologies, sparks new ideas through interdisciplinary dialogue, and paints a promising picture for intelligent healthcare. In the future, HUST will further expand cooperation dimensions, contributing "HUST wisdom" to the global advancement of healthcare intelligence.
Photos by：Rongjun Song
Written by：Haiyan Wang
