Under the Healthy China 2030 strategic plan, finding opportunities from historical nodes and exploring the intersection of medical care and technology will be the proposition of the times for the next decade.

”

Author | Liu Haitao

Editor | Li Yuchen

On the road to human health, medical treatment and medicine are the eternal panacea. With the implementation of the Healthy China Strategy and Healthy China 2030, the big health industry will lead a new round of economic development in my country.

Innovation in cutting-edge fields is the core force driving this wave.

How to find opportunities from historical nodes, explore the intersection of medicine and technology, and realize commercialization at the right time?

At this year's Medical Technology Summit, 19 industry leaders in medicine, production, research, and investment, respectively, from the two major tracks of medical imaging AI and AI pharmaceuticals, put forward their own insights for the development of the industry.

Academicians of four academies, chairmen of top international conferences, department heads of top tertiary hospitals, bigwigs from the investment community...both the level of guests and the diversity of topics are at the highest level in recent years.

The following is a review of the highlights of the conference:

1

AI Pharmaceutical Forum

Pan Yi: Using the third generation of artificial intelligence to assist biomedical big data research

As the opening guest of the morning session, Professor Pan Yi gave an opening speech entitled "Application of Artificial Intelligence in Biomedical Engineering".

He mentioned that biomedicine has entered the era of big data and AI, but there are many issues worth reflecting on: massive amounts of data are generated every day, but the level of data processing is not high. The main reason is that computer scientists do not understand biology, biologists do not understand programming, and the quality of the final output is relatively poor.

Professor Pan Yi believes: "For biologists, if they invest millions of dollars to improve research tools, but the final results are only slightly improved, the computer scientists who developed the tools will not be recognized."

Artificial intelligence technology can be applied to multiple stages of human life research and health management. Professor Pan focused on sharing the prospects and applications of the third-generation artificial intelligence technology. He said that the third-generation AI system combines the first-generation knowledge-driven and the second-generation data-driven to construct a more powerful AI. The key lies in knowledge embedding, multimodal data fusion and result interpretation.

Professor Pan Yi shared his latest research on "multimodal data fusion", which represents research entities such as genes, RNA, proteins, microorganisms, metabolites, pathways, and pathological and medical imaging data using networks at different levels, so that computational methods can be used to explore the potential relationships between biological entities.

Ren Feng: AI has great potential in the pharmaceutical industry where the failure rate is extremely high

The second speaker was Dr. Ren Feng, Chief Scientific Officer of Insilico Medicine, and his speech was titled "Artificial Intelligence Accelerates the Development of Fibrosis Drugs."

Ten days ago, Insilico Medicine's candidate drug with a new target and new molecular structure became the first AI-assisted innovative drug to enter the clinical trial stage in history.

In his speech, Dr. Ren Feng introduced in detail the research and development details of the IPF project, as well as the actual roles of Insilico Medicine’s three AI platforms: Chemistry 42, PandaOmics and InClinco.

Ren Feng said: "AI can do a lot of things in new drug research and development, such as being good at finding protein targets, or repurposing old drugs for known targets, and even using it in many 'quick-follow' projects. Everyone should have a more far-reaching idea about AI new drug research and development. The new drug molecules and new targets we think of today are just a small step, and we need to develop end-to-end next."

Tiannan Guo: AI+proteomics technology with over 90% diagnostic accuracy

After Ren Bo’s speech, Guo Tiannan, a distinguished researcher at West Lake University, gave a speech titled “AI-enabled proteomics big data to support precision medicine.”

Professor Guo’s speech is mainly divided into six parts:

First, what is proteomics?

Second, the latest technological advances in proteomics in clinical practice;

Third, the new concept of proteomics big data and the role of AI;

Fourth, AI assists in the diagnosis of thyroid nodules;

Fifth, AI can be used to classify COVID-19 in urine tests;

Sixth, a new concept of converting proteins into multi-dimensional matrices of Tensors.

Professor Guo said: "AlphaFold2 uses AI technology to make breakthrough progress in protein structure prediction, but the greater value of this research will be demonstrated in proteomics. Imagine a battlefield that requires different types of soldiers and weapons, and their performance is the protein structure. To win a battle, you need to know the number, performance, operation and repair methods of various soldiers and weapons, as well as the interaction of all military forces in the entire combat system. This process is proteomics. This is the relationship between protein structure prediction and proteomics."

At the end of his speech, Professor Guo also focused on the research on a new form of display of proteomic big data - how to convert proteomic data into a multi-dimensional matrix of Tensor.

"Tensors can be converted into videos. Each pixel is a polypeptide fragment of a protein. After flattening, we can get a regular image like the universe, which is distributed at intervals, and each interval is a molecular unit."

Song Le: Three major challenges facing AI new drug development

The fourth speaker of the morning was Song Le, chief AI scientist of Baidu Biotechnology and chairman of ICML 2022. The title of his speech was "Using Artificial Intelligence to Empower New Drug Research and Development."

Dr. Song Le mentioned that while everyone is looking forward to AI playing a huge role in new drug discovery, there are three issues that need to be considered in advance.

First, to have sufficient understanding of a disease, we need to have sufficient understanding of each organ in the body, the different functions of different cells, and the commonalities between cells. This will be a very complex network, requiring a lot of information such as the molecules accepted on the cell membrane, the mutual regulation in the cell, the protein gene expression in the cell, etc.

Second, we need to face the challenge of integrating data of different dimensions and scales, including gene sequencing, epigenetic group, protein expression, protein metabolism, tissue level, mechanism level, etc.

The third challenge is at the level of people. In the process of drug molecule or target discovery using AI models, data analysis and experiments are often done by two groups of people. Sometimes their ideas conflict, and sometimes communication takes a long time. There is a lack of a very efficient system that can integrate prediction, model output and experimental systems to accelerate iteration.

Roundtable Forum: AI pharmaceutical manufacturing is a technological revolution, and finding a closed loop is the key

The last session of the morning was the "AI New Drug Investor Discussion", which was a top discussion on the investment and financing of AI new drug research and development and the next step of development in China. It was hosted by Qin Zhen, Executive Director of Ali Health Investment Department. The four top investors attending the roundtable were:

Yang Kun, partner of Gaorong Capital, Zhou Yi, executive general manager of Shenzhen Capital Group and general manager of the health industry fund investment department; Liu Mingyu, general manager and founding partner of Bangqin Capital.

Zhou Yixian, executive general manager of Shenzhen Capital Group, expressed his views on the development status and capital trends of the AI new drug research and development industry.

He believes that AI pharmaceutical manufacturing has just started, and the molecules or targets discovered through AI pharmaceutical manufacturing technology have not yet been clinically verified. If more subsequent cases can be run successfully, people's trust and dependence on AI will increase. "I don't want AI pharmaceutical manufacturing to be like AI medical imaging, where big ups will inevitably be followed by big downs. For entrepreneurs, AI pharmaceutical manufacturing is a good choice, but don't be too anxious. Making medicine is slow, and there are still many hurdles to overcome."

Liu Mingyu, general manager of Bangqin Capital, believes that "if the new tool is a technological revolution, it is possible to subvert the traditional rules of the game. AI pharmaceuticals still need certain breakthroughs to verify the difference from traditional ways of thinking, but AI pharmaceuticals have stronger 'tool' attributes."

During the roundtable discussion, the four investors also discussed the high risks of AI pharmaceutical manufacturing.

In this regard, Yang Kun, partner of Gaorong Capital, said that the prospects and risks of AI pharmaceuticals need to be viewed from the perspective of the closed loop of the industry. Taking AI diagnosis as an example, it has a practical role in clinical practice, but its commercialization performance varies in China and the United States. The touchstone of AI pharmaceuticals will come faster.

Currently, many AI-developed drugs are in the preclinical stage. Once they enter the clinical stage, they will face two problems. First, entering the clinical stage means that the company will enter the valuation system of new drug companies; second, the effect of molecules screened by AI compared with those studied by scientists remains to be verified. New drug development naturally has a certain failure rate, which will also have a certain impact on the industry and companies. In the next two years, AI pharmaceutical companies may usher in a differentiation between "going up" and "landing".

Qin Zhen, executive director of Alibaba Health Investment Department, concluded: Biocomputing will move more from static prediction to dynamic direction. AlphaFold2 is a screenshot of the conformation of a three-dimensional structure. In the future, expectations for it will change from photos to videos, and we will truly see how proteins move.

In addition, from the prediction of proteins to the secondary structure of RNA, and now also the tertiary structure, connecting its structure and movement together is also a trend.

Third, the combination of dry and wet experimental data requires a continuous closed loop, with new real experimental data that are then fed back into the algorithm. This is also the AI pharmaceutical trend that everyone expects to see.

2

Medical Imaging AI Forum

Xiao Yi: Medical imaging AI has achieved a head effect, and three values will be verified

In the opening report of the Medical Imaging AI Sub-Forum, Professor Xiao Yi from the Department of Imaging at Shanghai Changzheng Hospital delivered a speech entitled "Variables and New Driving Forces in the Development of Medical Imaging AI".

Xiao Yi said that with the increase in understanding of disease diagnosis and treatment, medical imaging artificial intelligence products are further integrated horizontally and vertically to develop in depth and breadth, and complete solutions are the good products that can truly serve the clinic.

Director Xiao took several leading AI companies in the industry as examples and shared their layout in clinical and scientific research directions.

However, medical AI still faces commercial challenges. "The leading AI companies are still suffering from the pain of bleeding and are unable to enter the medical insurance market. Even if the leading AI companies have completed IPOs, they are still in a dilemma of a broken business chain with only bleeding, no transfusions, and no blood recovery."

However, looking to the future, under the background of hierarchical diagnosis and treatment, the service demand of primary medical institutions will grow rapidly. At the same time, as the three types of certificates are issued one after another and the leading companies fully demonstrate their good clinical, economic and social value, medical AI companies will usher in new entrepreneurial propositions and a new growth cycle.

Zhang Xiaochun: The shared medical system in Fangcang is the general trend

Director Zhang Xiaochun is the head of the Imaging Department of Guangzhou Women's and Children's Medical Center, and is currently the chairman of the Fangcang Medical Branch of the China Association for Medical Devices Industry.

She said that the innovation of new technologies such as 5G+AI+Brain-Computer Interface has brought about tremendous changes in the current medical production relationship. Director Zhang Xiaochun focused on sharing the construction achievements of the "square cabin sharing" medical model.

She believes that in the post-epidemic era, what kind of medical model will medical practitioners adopt? It must be to establish a mobile intelligent shared integrated functional carrier with a certain function or a combination of multiple related functions, that is, a square cabin shared medical system, which is the general trend. Doctors, scientists, entrepreneurs, and government functional managers must have the mentality of using all tools at their disposal and have the ability and confidence to master advanced intelligent tools.

Yuan Jin: In the past, we talked about artificial intelligence in ophthalmology, but in the future we will talk about intelligent ophthalmology

As the third speaker of the sub-forum, Professor Yuan Jin, deputy director of Zhongshan Eye Center, has actively led his team in recent years to complete the design and evaluation of new ophthalmic imaging equipment such as ultra-high-resolution OCT and ophthalmic multimodal imaging systems, as well as the development of intelligent analysis technologies such as ophthalmic artificial intelligence diagnostic cloud systems.

In his speech, Professor Yuan Jin said that the development of AI products has three elements: algorithms, computing power and data, and data is crucial.

In order to facilitate product development for researchers, Sun Yat-sen University Zhongshan Eye Center released the gold standard fundus dataset to the world for free, named iChallenge. Currently, more than 2,000 teams around the world have used this dataset to develop clinical AI applications.

Professor Yuan believes that the industry used to talk more about artificial intelligence in ophthalmology, but now this concept should be expanded to "smart ophthalmology": using artificial intelligence, as well as new-era technologies such as 5G, wearables, high-definition imaging technology, and a new generation of robots, integrated into clinical applications to create true smart ophthalmology, rather than just being limited to the concept of artificial intelligence in ophthalmology.

Currently, Professor Yuan’s team is working with Shenzhen Rui Medical to promote an all-weather, multi-scenario, interactive visual function navigation system to create a new intelligent vision assistance system for people with low vision and blindness to change the quality of life of this group of people.

Li Yuwei: The exact path for the future commercialization of medical AI products

The fourth speaker in the afternoon was Li Yuwei, chief scientist of CoYa Medical. The title of his speech was "Starting from Clinical Needs - The Road to Commercialization of AI Products."

During the morning speech, an investor mentioned the commercialization difficulties of medical imaging AI. As the first company to obtain Class III certification for medical imaging AI products in my country, Coa Medical is also one of the pioneers that has done the most commercial exploration of medical AI products. In his speech, Dr. Li Yuwei introduced the research and development, clinical verification and compliance implementation, price approval and commercialization model of medical AI products.

Dr. Li Yuwei said that as a leading company in the industry, looking back on the development of Coa Medical in the past few years is actually very similar to the evolution history of medical AI as a whole. I have two feelings: one, it is not easy; two, I am lucky.

"We developed the core algorithm of Deep Pulse Score as early as 2016, and began clinical trials at the end of that year. Then, we entered the special approval channel for innovative medical devices of the State Food and Drug Administration in 2018, and were officially approved in early 2020. The time period for implementation was far beyond our expectations as researchers;"

"For the same reason, we are fortunate to be the first company approved to fill the gap in the implementation of medical AI product supervision in China. Starting from scratch, we have worked with regulatory authorities, especially the State Food and Drug Administration, to explore how to implement medical AI products in my country. This has enabled Keya to accumulate profound experience in the commercialization of medical AI products and understand how to conduct targeted product design, data and algorithm research, and clinical application around the commercialization of medical AI products from the very beginning."

Huang Feng: Traditional imaging equipment companies must learn how to “build roads” and “open to traffic”

The fifth speaker in the afternoon was Dr. Huang Feng, chief scientist of Neusoft Medical Systems. The title of his speech was “Exploring Paths to Help Solve Medical Pain Points with AI.”

Dr. Huang Feng said that when people think about the artificial intelligence applications of traditional imaging equipment companies, the most common concepts are automatic positioning, rapid reconstruction, low-dose CT imaging, and other artificial intelligence applications surrounding the equipment, but how can they further meet the clinical needs of users?

Based on the current issues and status quo of the implementation and commercialization of medical AI, insufficient and uneven distribution of medical resources, Neusoft Medical has developed a platform, MDaaS, dedicated to connecting medical devices and medical imaging data to serve all parties in the medical ecosystem and integrate all stakeholders from equipment to medical institutions, governments, research institutes, patients, etc.

Specifically, Dr. Huang Feng believes that MDaaS mainly does two things: building roads and opening them to traffic.

Building roads means that MDaaS has built platform services for several different scenarios internally, including NeuMiva, an intelligent imaging public cloud platform for grassroots medical institutions, e-Stroke, a stroke platform for intelligent diagnosis and tiered treatment of specialized diseases, and eLungCare, a lung disease platform.

Opening to traffic means achieving interconnection of the ecological chain, connecting equipment, medical service providers, patients, governments, scientific research institutions and third-party service providers, realizing data interoperability and the application of artificial intelligence products.

IEEE Fellow Roundtable Dialogue: Conducting high-level academic research and a dialogue mechanism between industry, academia and research

The IEEE Fellow roundtable dialogue of the Medical Imaging AI Forum was co-organized by Leifeng.com and the International Symposium on Image Computing and Digital Medicine (ISICDM). It was hosted by Dean Feng Qianjin of the School of Biomedical Engineering of Southern Medical University, and the guests attending were:

Jiang Tianzai, researcher at the Institute of Automation, Chinese Academy of Sciences, and IEEE Fellow;

Shaohua Zhou, Professor of the University of Science and Technology of China and IEEE Fellow;

Li Chunming, professor of School of Electronic Engineering, University of Electronic Science and Technology of China, and IEEE Fellow;

Peng Hanchuan, founding dean of the Institute of Brain Science and Intelligence Technology of Southeast University and IEEE Fellow;

Zheng Yefeng is the director of Tencent Tianyan Lab and IEEE Fellow.

During the dialogue that day, the six guests' discussion mainly focused on two aspects: academic research and the integration of industry, academia and research:

Regarding the academic level. Professor Jiang Tianzai believes that everyone wants to do high-level scientific research, but what is high-level scientific research is a process of cognition. Everyone may have different standards, and the same person may not have the same standards at different times. Professor Jiang Tianzai majored in mathematics from undergraduate to doctoral degree. In the ten years from 2000 to 2010, he was basically engaged in clinical basic research. He published some high-level papers, but they could not solve clinical real problems. When applying for the 973 project, it is determined to take clinical needs as the starting point and conduct research that is meaningful to the discipline and society. Only such papers will reflect their true level and value.

Professor Zhou Shaohua approved of this idea. He said: "When I was working at Siemens Healthineers, I was fortunate to get into this kind of research by accident. Many of the problems I worked on came from clinical and academic research on medical imaging. It is a better way to start from clinical problems. Clinical problems are the source of many important problems."

Professor Li Chunming also believes that "technical research on medical imaging cannot blindly rely on a single technology. For example, there has been a lot of research on deep learning, and it has been well applied in many fields, but deep learning cannot solve all problems. Ultimately, it depends on whether it can solve practical problems, meet clinical needs, and truly help doctors improve their work efficiency and benefit patients."

The second topic discussed by the scholars was how to achieve integration of industry, academia and research.

Professor Peng Hanchuan believes that many of the practical pain points of doctors can be solved by engineers. For example, doctors need to record the data samples or clinical samples they come into contact with, which requires some convenient annotation tools. Therefore, more than a decade ago, Professor Peng Hanchuan developed a set of tools that allow doctors to quickly and effectively annotate three-dimensional data.

As the head of Tencent Tianyan Lab, Dr. Zheng Yefeng also introduced his experience in developing a minimally invasive aortic valve replacement surgery navigation system from an industrial perspective. He said that it is very difficult to accurately find the valve position without X-rays and contrast agents. It was only after communicating with doctors, and even wearing lead aprons to observe in the operating room, that he came up with the idea of building a navigation system.

Finally, Dean Feng Qianjin also made a summary. He believed that different viewpoints need to be fully expressed, which is the value of this forum. Medical image analysis is a cross-border job, which requires researchers to start from practical clinical problems and continuously optimize methods and results. It is believed that the academic experience of the five IEEE Fellows will also be able to point out a clear growth path for young scholars in China.

More details of the forum will be presented in the next two weeks.

END