At the 2021 World Artificial Intelligence Conference, ZHU Wentao, Deputy Director of the Zhejiang Lab Research Center for Healthcare Data Science, was honored as one of the Top 10 Distinguished AI Professionals in the Yangtze Delta.

ZHU Wentao holds a bachelor's degree from Tsinghua University and a PhD degree from the University of Southern California. Focusing on intelligent medical imaging, he is the first Chinese winner of the Arthur Weis Award, an important award presented by the Society of Nuclear Medicine and Molecular Imaging.

As one of the earliest research scientists to join Zhejiang Lab, ZHU's research path is integrated with the Lab's exploration for institutional innovation. His project on intelligent PET/CT imaging systems sets a new standard for industry-academic-research integration and innovation.

A New PET/CT System Based on Intelligent Computing

PET/CT is one of the most cutting-edge techniques in modern medicine. It has been widely used in areas such as tumor screening and the diagnosis of neurological and cardiovascular diseases. However, the medical community still remains cautious about PET/CT scans due to radioactivity issues. The clinical radiation dose of a PET/CT scan is about 7.5 mSv, while the radiation received by a person in the natural environment is only about 2.4 mSv per year.

One of the core technologies of Zhejiang Lab's intelligent PET/CT medical imaging system is "low-dose PET imaging". It produces PET images for clinical diagnosis while reducing the radiation dose by 50% for body scans and 70% for head scans. "Our team has developed a deep learning imaging algorithm based on PET raw data. Combined with a physical model of PET imaging reconstruction, our system can directly process the raw data of the imaging device and thus greatly reduce the loss of effective information. We can therefore obtain clearer PET images and a stronger ability to detect small lesions," says ZHU. In addition, the team has developed a fully automated steering and positioning system for tuning left ventricle PET images to the standard view for clinical analysis. The system has attained a 100% success rate in nearly 100 cases for clinical validation. It matches more than 95% of clinicians' operations, with a quantization error of less than 5%. Unlike traditional automatic steering algorithms, the algorithm designed by ZHU's team eliminates the need for pre-processing tasks such as left ventricular segmentation, which largely increases the speed of steering operations. Moreover, the system is operated according to global information, making it less dependent on the structural integrity of the left ventricle, and thus achieving a higher success rate in clinical practice.

These significant achievements come from innovative attempts that extend intelligent computing to the field of medical imaging. ZHU's team has published a number of papers in prestigious journals related to medical imaging, such as The European Journal of Nuclear Medicine and Molecular Imaging and The Transactions on Radiation and Plasma Medical Sciences. The team is expected to continue in its development of AI-empowered medical imaging devices.

 A Quickly Growing Team with the Support of the Lab

It took only one and a half years for ZHU's team to obtain their milestone achievements. ZHU believes that their success should be attributed to the philosophy of making the best use of individual strengths.

"We are engaged in interdisciplinary research activities that are strongly application-oriented. Our work requires a system of knowledge that includes multiple disciplines in the field of information, medicine, and engineering. A specific area requires a specific person with matching knowledge and skills. Each member of our team has unique expertise that is indispensable to the team." According to ZHU, this is the key competence of his team.

But in the eyes of his team members, ZHU himself has the most comprehensive knowledge. "In the field of medical imaging, most people are only familiar with one technology. But Dr. ZHU has in-depth research experience in multiple areas, such as image reconstruction, image registration, and image analysis. Therefore, he can guide the project from a more macro and profound perspective. That is one of the reasons why our technological outcomes can be ready for real-life application so quickly," says RAO Fan, a member of ZHU's team.

In addition, ZHU's persistent dedication to his scientific pursuit has also motivated the whole team to keep moving forward. With a vision of making China better through scientific research, ZHU joined the start-up team of Zhejiang Lab at the beginning of its establishment. "For me, the most significant part of being a science professional is doing research that fulfills the needs of society, and using my technologies to solve real-life problems," says ZHU.

With support from the Lab's leadership team, ZHU quickly organized a research team and started his project. Moreover, he and his team produced high-level research outcomes within a quite short period of time. "The Lab has a flexible mechanism for talent recruitment and appointment. It also gives full trust and support to our researchers. That's why we can achieve our goals at such an amazing speed," says ZHU.

A New Mechanism for Industry-Academia-Research Cooperation

In addition to academic outcomes, the technologies developed by ZHU and his team have already been used in PET/CT devices made by a major manufacturer of imaging equipment in China. These devices have been used in clinical applications. This is the first successful research-industry cooperation project at the Lab. From research to commercialization, the entire process only took about one and a half years.

The commercialization process of most research projects follows a step-by-step procedure, in which the research and development of a technology comes before cooperation with the industry. For Zhu's project on intelligent PET/CT medical imaging, the cooperation took place at the beginning of the research. When the project only had a conceptual framework, it received financial support from MinFound Medical Systems to develop cutting-edge technologies.

The Lab and MinFound soon began their concrete research activities, and carried out industry-academia-research cooperation at a quick pace. "Our technologies were all inspired by the need for high-performance medical imaging devices and the real pain points in hospital scenarios," says ZHU. According to him, the direct link with the industry gave the team clearer research goals. In addition, as the team was task-oriented, members would work together to discuss and tackle problems, significantly accelerating the commercialization process.

Next, ZHU's team will continue to develop an intelligent medical imaging software system for industrial use. The team will also cooperate with well-known hospitals for clinical research and application. Zhejiang Lab has been deepening the mechanism for industry-academia-research cooperation and practicing new approaches in the process of working with industry, so as to develop the best mechanism for the Lab.