Recently, the team led by Professor Huang Han from our school, in collaboration with the team led by Professor Ling Fei from the School of Bioscience and Engineering and the team led by Professor Yao Xueqing from Guangdong General Hospital, has achieved a major scientific research breakthrough. They have successfully developed the world's first artificial intelligence model that can accurately predict the recurrence risk after local resection of colorectal cancer. This achievement has been published in the official journal of the European Society of Surgical Oncology, European Journal of Surgical Oncology, opening up a brand-new path in the field of precision treatment of colorectal cancer.As soon as this groundbreaking achievement was unveiled, it attracted widespread attention from all sectors of society. Many authoritative media outlets, including Xinhua News Agency, China Daily website, The Paper, and Southern Metropolis Daily, have conducted in-depth reports on this achievement, quickly sparking a heated discussion in the medical and scientific communities.
Screenshot of the Thesis
Colorectal cancer ranks third in terms of incidence among all cancers globally, accounting for approximately 10% of all cancer cases. For patients with early-stage colorectal cancer, local resection is a commonly used treatment method. However, the problem is that some patients with a high risk of recurrence need to undergo additional surgery after the initial operation. Moreover, current medical methods find it difficult to accurately determine the recurrence risk, leading to frequent occurrences of both overtreatment and undertreatment. In this severe situation, a clinical decision-making assistance tool that can accurately predict the recurrence risk after surgery has become a crucial issue to be solved in the field of colorectal cancer treatment.
The scientific research cooperation team took an alternative approach and focused on the pathological images of T1-stage colorectal cancer specimens resected by endoscopic or transanal surgery. They applied technologies such as medical image data augmentation algorithms and deep residual networks to keenly capture 107 key pathological features in the tissue images. Microstructural features that are difficult to distinguish with the naked eye, such as the depth of tumor invasion and vascular invasion, cannot escape the sharp eyes of this technology. The team also innovatively constructed a three-dimensional feature space mapping to accurately identify the distribution pattern of abnormal cells at the 0.5μm level in traditional HE-stained slices. With an advanced algorithm framework based on transfer learning, this model demonstrated an astonishingly high accuracy rate of 97.9% in the laboratory test environment.
The successful development of this AI model not only provides clinical doctors with a solid and reliable basis for decision-making but also strongly proves that artificial intelligence technology based on tissue images has great potential and excellent performance in predicting tumor behavior. Professor Huang Han and Professor Ling Fei jointly established the AI-for-Oncology Group. Currently, they are working closely with Nanfang Hospital of Southern Medical University and are fully committed to the research and development of an AI prediction model for the recurrence risk after resection of hepatocellular carcinoma using HE-stained slices, and have already achieved phased results.
The Team Led by Professor Huang Han Participated in an International Academic Conference
Our school closely follows the strategic layout of interdisciplinary subjects of the university, encourages teachers to carry out interdisciplinary research, and promotes interdisciplinary cooperation in Artificial Intelligence and Big Data + X. The successful development of this AI model is a typical achievement of interdisciplinary cooperative research. Due to its practicality and innovation, it has also attracted the attention and reports of authoritative media at the central, provincial, and municipal levels, such as Xinhua News Agency, China Daily website, and Southern Metropolis Daily.
