"EchoCare" Ultrasound Foundation Model Launched in Hong Kong as CAIR's New Breakthroughs Propel AI Ultrasound into the Era of Large Models
HONG KONG, Sept. 17, 2025 /PRNewswire/ -- Amid the global surge in rapid advancements in artificial intelligence, the healthcare sector is entering a critical phase of intelligent transformation. As a vital tool in clinical diagnosis, ultrasound imaging has long been plagued by issues such as low efficiency, inconsistent diagnostic standards, and insufficient AI model accuracy. These challenges urgently call for technological breakthroughs and industrial collaboration. Against this backdrop, Centre for Artificial Intelligence and Robotics (CAIR), Hong Kong Institute of Science & Innovation (HKISI), Chinese Academy of Sciences unveiled its latest scientific achievement – the "EchoCare" Ultrasound Large Model – on 17th September in Hong Kong.
At the press conference, Prof Hongbin Liu, Director and Professor of CAIR, Prof Gaofeng Meng, Deputy Director and Professor of CAIR, Prof Jiebo Luo, Member of Academia Europaea and the US National Academy of Inventors, Prof Randolph Hung-leung Wong, Chief of the Cardiothoracic Surgery at The Chinese University of Hong Kong (CUHK), were joined by numerous renowned scholars, clinical experts, and more than ten media representatives to witness this milestone breakthrough in AI-powered ultrasound medicine.
Pioneering Structured Contrast Self-Supervised Learning Framework
The "EchoCare" Ultrasound Large Model was trained on the first ultrasound image dataset known to exceed 4 million images. The model introduces a "Structured Contrast Self-Supervised Learning Framework", which leverages hierarchical tree labels derived from medical priors to enable multi-label semantic relational structured learning and implicit encoding. Through techniques such as Masked Image Modelling (MIM), Adaptive Hard Patch Mining, and Progressive Training, the model effectively enhances its ability to model the deep semantic features of ultrasound images and improves generalisation performance.
Test results demonstrate that "EchoCare" achieves state-of-the-art (SOTA) performance across seven medical tasks, including image segmentation, classification, detection, regression, and enhancement, as well as in over ten downstream applications. On average, it delivers a 3%-5% improvement compared with current SOTA methods.
A Major Milestone for Inclusive Smart Healthcare
In his opening remarks, Prof Jiebo Luo congratulated the successful development of the "EchoCare" Ultrasound Large Model and highly praised it as another significant breakthrough in the deep integration of AI and medical applications. He noted that the implementation of "EchoCare" in routine hospital examinations can significantly reduce reliance on medical specialists while assisting doctors in making diagnoses more efficiently and accurately. This technology is expected to substantially improve the efficiency of medical services while providing greater opportunities for the optimal allocation of healthcare resources.
"Listening to Sound, Grasping Principles"
At the press conference, Prof. Gaofeng Meng, Deputy Director of CAIR, explained that the name "EchoCare" originates from the idiom "Listening to Sound and Grasping Principles" (Ling Yin Cha Li) in Liu Xie's The Literary Mind and the Carving of Dragons (Wenxin Diaolong • Zhiyin). The text notes that "One becomes proficient in music only after mastering a thousand melodies, and gains insight into weaponry only after observing a thousand swords." which aligns closely with the philosophy behind developing the ultrasound large model. He further introduced that EchoCare innovatively adopts a structured contrastive self-supervised learning approach, enabling feature learning and downstream task decoupling without the need for extensive data annotation, thereby achieving the internalization of prior knowledge in the ultrasound domain and facilitating cross-task knowledge transfer.
Prof. Meng emphasized that, unlike traditional large models, EchoCare innovatively adopts a structural self-supervised learning approach. It removes the need for extensive data annotation, enables feature learning, and decouples downstream tasks, thereby internalizing prior knowledge in ultrasound and facilitating cross-task knowledge transfer.
He also showcased the model's technical highlights, data advantages, and application results. Specific case validations included 1,556 ovarian tumor ultrasound cases at Qilu Hospital of Shandong University and more than 1,000 thyroid ultrasound examinations at Xiangya Hospital of Central South University, where EchoCare significantly outperformed existing state-of-the-art methods.
Relieving Physicians, Benefiting Patients, and Offering Immense Clinical Value
The standardized analytical capabilities of "EchoCare" can effectively reduce the rates of missed and misdiagnosed major diseases, significantly enhancing the efficiency and standardisation of clinical diagnosis. It provides robust technical support for frontline medical practitioners.
During the case-sharing segment, Prof Randolph Hung-leung Wong from CUHK first presented retrospective validation results for the detection and analysis of aortic aneurysms using "EchoCare" in cardiac ultrasound. He also envisioned the potential clinical value of integrating this large model with robotic technology. In the live demonstration, he showcased two ultrasound scanning videos, where the model rapidly captured and analysed the key medical information, successfully identifying abnormal cases and automatically generating ultrasound reports for the doctors' reference.
Various Views: Technology for the People, AI Deeply Rooted in Reality
During the media Q&A session, Prof Hongbin Liu, Prof Gaofeng Meng and Prof Randolph Wong answered questions from media representatives, such as Phoenix TV. They engaged in in-depth discussions on the technical details, clinical applications, and commercialisation pathways of the EchoCare Ultrasound Large Model. Following the Q&A, the media representatives conducted exclusive interviews with the invited guests and visited the Embodied-AI Surgery Platform developed by CAIR, where they experienced CAIR's latest AI healthcare achievements up close.
The "EchoCare" Ultrasound Large Model, open-sourced by the Centre for Artificial Intelligence and Robotics (CAIR), Hong Kong Institute of Science & Innovation, Chinese Academy of Sciences, breaks down the compatibility barriers between traditional ultrasound devices and unlocks the value of multi-centre data, providing medical institutions with reusable AI infrastructure. This achievement not only accelerates the large-scale deployment of ultrasound AI but also injects sustained innovation into the advancement of the smart healthcare industry.
Released by
The Centre for Artificial Intelligence and Robotics (CAIR).
Established in 2019, the Centre for Artificial Intelligence and Robotics (CAIR) is one of the two centres under Hong Kong Institute of Science & Innovation (the only directly affiliated research institute of Chinese Academy of Sciences in Hong Kong).
CAIR is dedicated to integration and innovation of artificial intelligence and life sciences, conducting research in three main areas: Multimodal AI Large Model, Embodied Intelligent Robots, and Intelligent Sensing Technologies. CAIR is a key institution supported by Hong Kong's InnoHK initiative in the field of AI. It is among the few institutions globally that systematically carry out research and development of AI systems for medical and healthcare applications, as well as their technological transformation.
SOURCE CAIR
Share this article