Union Hospital Affiliated with Huazhong University of Science and Technology to Transfer Biliary-Pancreatic Ultrasound Recognition Technology for RMB 2 Million
Recently, Union Hospital, Tongji Medical College of Huazhong University of Science and Technology, released a public notice on the transformation of scientific and technological achievements. The hospital intends to transfer its intellectual property through negotiated pricing.“A Scanning Identification System and Method for the Normal Anatomical Structures of the Biliary-Pancreatic System”Relevant patents are transferred to industry partners for use, with a total transfer amount of RMB2 million yuan. The inventor of this patented technology isLin Rong。
Lin Rong:Chief Physician and Deputy Director of the Department of Gastroenterology at Union Hospital, Tongji Medical College, Huazhong University of Science and Technology; concurrently serving as Chief Expert and Department Director of the Department of Gastroenterology at the Chegu Campus of Union Hospital. She has long been engaged in clinical practice and scientific research on digestive tract diseases, focusing on early screening for gastrointestinal cancers, diagnosis and treatment of tumors, and atrophic gastritis. She has presided over multiple national-level scientific research projects, with cumulative research funding exceeding RMB 8 million. As Vice Chairperson of the Youth Committee of the Chinese Society of Gastroenterology, Deputy Leader of the Working Group on Gastrointestinal Hormones and Neuroendocrine Tumors, Member of the Helicobacter pylori Working Group, and Member of the Oncology Collaboration Group, she holds significant influence in the academic community. Since 2023, she has served as a member of the Gastroenterology Group of the Medical Emergency Expert Team under the National Health Commission, leading the development of the Department of Gastroenterology at the Chegu Campus into a key clinical discipline and advancing"Integrated Management, Consistent Quality Healthcare"。
This technology provides an ultrasound imaging scanning and recognition system and method targeting the normal anatomical structures of the biliary-pancreatic system, with its core purpose being to address issues in ultrasound examination of the biliary-pancreatic system“Difficulty in identification, incomplete scanning, reliance on experience”and other issues.
Leveraging technological means to assist ultrasound operators (such as physicians and trainees) in rapidly and accurately identifying biliary and pancreatic images acquired via linear array endoscopic ultrasound, thereby ensuring a comprehensive scanning process and reliable results while reducing reliance on individual operator experience.
Ultrasound Scanning of the Biliary and Pancreatic Systems(Especially linear-array endoscopic ultrasound) serves as a core method for early screening and lesion localization in biliary and pancreatic diseases, playing a critically important role in the diagnosis of conditions such as pancreatic cancer, bile duct stones, and chronic pancreatitis. Linear-array endoscopic ultrasound can penetrate the gastric wall to directly scan the biliary and pancreatic systems, clearly visualizing subtle structural details of these organs. Furthermore, it provides guidance for fine-needle aspiration biopsy of suspicious lesions, which directly influences the formulation of treatment plans and patient prognosis, making it an indispensable component of clinical diagnosis and management.
Clinically, both novice physicians and experienced senior practitioners encounter numerous practical challenges when performing ultrasonographic scanning of the biliary and pancreatic systems.
For beginners, the numerous organs surrounding the biliary and pancreatic systems, coupled with their intricate anatomical structures, present a significant challenge. Moreover, ultrasound images are predominantly single-plane. Such images are highly sensitive to spatial structural features, requiring physicians to accumulate extensive experience over a prolonged period to accurately identify the subdivided structures of key areas, including the gastric region, the duodenal bulb, and the descending part of the duodenum. Thus, the learning curve is notably steep.
Even experienced physicians are highly susceptible to various interfering factors during scanning. For instance, unstable image quality, variations in individual anatomical structures, and differences in subjective judgment can all lead to errors in structural identification, thereby adversely affecting the accuracy of the examination.
The current lack of standardized intelligent assistance tools in clinical scanning further exacerbates existing pain points. On one hand, traditional scanning relies on physicians’ subjective judgment and lacks unified standards for structural recognition. There is an absence of effective methods for extracting spatial structural features, which carry greater weight in single-channel ultrasound images, leading to missed or misidentified subtle structures and increasing the risk of misdiagnosis. On the other hand, the scanning process lacks systematic documentation and guidance, making it difficult for physicians to determine in real time which areas have been scanned and which remain uncovered, thereby resulting in incomplete examinations. Meanwhile, existing technologies struggle to standardize the storage of intermediate data and operational steps during the scanning process, hindering subsequent traceability analysis and workflow optimization.
Furthermore, traditional ultrasound image recognition techniques have inherent limitations. Most existing image processing methods based on convolutional neural networks are designed forMulti-channel Imaging Design Featuring Highlight Maps, which fails to meet the feature extraction requirements for single-channel ultrasound imaging, thereby resulting in low accuracy of structural recognition.
These challenges make ultrasound scanning of the biliary and pancreatic system technically difficult and poorly consistent. Some patients face risks of overtreatment, undertreatment, or delayed diagnosis due to inaccurate or incomplete examination results. Therefore, there is an urgent need for an intelligent assistance system capable of accurately identifying anatomical structures and standardizing scanning protocols to address this industry-wide dilemma.
This scanning and identification technique for the normal anatomical structures of the biliary and pancreatic system, utilizing“Intelligent Algorithms + Standardized Processes + Full-Cycle Assistance”'s innovative design precisely addresses the pain points of traditional scanning, presenting three core advantages.
In terms of technological innovationAdopting a Deep Learning Architecture of “Convolutional Neural Network + Attention-Enhanced Model”, specifically addressing the challenge of recognizing spatial structural features to which single-channel ultrasound imaging is sensitive.
During the preprocessing phase, leveragingImage Scaling, Cropping, and Adaptive Histogram Equalization Algorithms, uniformly convert the original images into single-channel images with the same resolution, laying the foundation for precise recognition.
In the core recognition phase, after convolutional neural networks (ResNet or VGGNet) extract multi-dimensional convolutional feature tensors, the attention-enhanced model strengthens focus on high-weight spatial structural features by calculating channel weights and fusing the results of average pooling and max pooling, thereby significantly reducing the probability of misclassification.
Furthermore, the system supports the integration of multiple attention-enhancement modules at intermediate or terminal stages of the convolutional neural network, further optimizing feature extraction. This enables more precise identification of 12 subdivided regions of the stomach, 4 regions of the duodenal bulb, and 3 regions of the descending part, thereby thoroughly overcoming the limitations of traditional algorithms in the accuracy of ultrasound image recognition.
Technology builds up"Acquisition - Recognition - Feedback - Guidance"closed-loop process, assisting the operator in completing the scan in a standardized manner throughout. The imaging display module synchronously presentsOriginal Ultrasound Images, Recognition Results, and 3D Stereoscopic SimulationsIf the results meet the criteria, the procedure can be concluded directly; if not, a prompt to continue data acquisition will be issued to prevent process interruption or omission due to subjective judgment. The database stores intermediate status data, operational steps, ultrasound images, and recognition results in real time during the scanning process, enabling full traceability of the entire procedure and facilitating the optimization of scanning standards through data accumulation.
Whether for inexperienced beginners or seasoned physicians, this technology enables rapid mastery of key scanning points, reducing the impact of individual experience variability on results. Meanwhile, the unscanned area alert function ensures that all critical regions are covered, significantly enhancing the completeness of the scan.
Addressing the pain points of traditional scanning, namely the "long learning curve and reliance on experience," this technology significantly shortens the onboarding time for beginners through intelligent assistance. Operators no longer need to spend extensive time accumulating experience in anatomical structure recognition; instead, they can quickly and accurately locate target areas by following real-time identification results and 3D simulation guidance. Furthermore, standardized operational procedures and clear criteria for result interpretation lower the professional skill requirements for operators, enabling the widespread application of this technology in ultrasound examinations of the biliary and pancreatic systems across medical institutions at all levels.
Meanwhile, this technology is compatible with the clinical application scenarios of linear array endoscopic ultrasound (EUS). It not only provides precise support for screening early-stage biliary and pancreatic diseases but also offers accurate guidance for puncture biopsy of suspicious lesions, effectively avoiding overtreatment or undertreatment while balancing clinical practicality and universality.
The current AI-assisted ultrasound diagnostics market has established a competitive landscape characterized by “international giants + domestic technology enterprises + innovative startups.” Competing products primarily focus on ultrasound image recognition and assisted diagnosis and treatment for the digestive system (including the biliary and pancreatic systems), with core technical pathways centered on optimizing deep learning algorithms and adapting to multi-scenario equipment. However, there remains room for differentiated breakthroughs in the precise identification of detailed anatomical structures within the biliary and pancreatic systems and in providing standardized guidance for scanning procedures.
NorthJingshi Perception Intelligence Technology Co., Ltd.,Founded in 2023, the company focuses on the niche sector of AI-powered ultrasound. It offers the Zhiying Musculoskeletal Ultrasound AI-Assisted Diagnosis and Treatment System (with extended coverage to certain digestive system scenarios) and has completed multiple rounds of financing. Its core product“Zhiying”Centered on musculoskeletal ultrasound, it also deploys AI modules for thoracic and abdominal ultrasound, enabling real-time dynamic visualization of anatomical structures, standard plane referencing, and anatomical guidance.
This product has been adapted to mainstream high-end ultrasound systems as well as domestically produced portable ultrasound devices, and has been implemented in over 100 hospitals across multiple departments, including pain management and rehabilitation. Currently, “Zhiying” is expanding its applications in digestive system ultrasound, having initially achieved structural recognition of organs such as the gallbladder and pancreas.
Chao Yan Co., Ltd.(Super Research Co., Ltd.)’s AI-powered diagnostic platform for full-volume ultrasound has garnered significant attention. As an innovative enterprise focused on the field of AI-enabled ultrasound, Super Research’s core technological advantages are primarily embodied in its closed-loop solution of “automated scanning + AI analysis.”
This full-volume ultrasound AI diagnostic platform enables standardized data acquisition through its patented probe and coupling pad design. Scanning can be successfully performed even without the involvement of senior physicians. Meanwhile, the AI system simultaneously conducts image reconstruction and lesion screening.
Currently, the platform has successfully obtained NMPA certification and EU CE marking. By 2024, it aims to achieve coverage in 80% of relevant departments at Grade III Class A hospitals across China.