The conference of AI medical device innovation and cooperation platform was held in Boao. Read and understand the common problems of AI medical device review and approval

In July 2019, in order to build an open and collaborative innovation system of AI medical devices and promote the evaluation and approval of AI medical devices, the AI medical device innovation cooperation platform came into being Half a year, the AI medical device innovation cooperation platform has made a series of achievements In the end of 2019, the "2019 AI Medical Device Innovation Conference and AI medical device" sponsored by the AI medical device innovation cooperation platform, China Academy of information and communication, supported by the national pharmaceutical regulatory and medical device technology evaluation center, and organized by Hainan drug regulatory authority, Hainan Boao Lecheng International Medical Tourism pioneer area authority and Boao Ecological Software Park Machinery innovation cooperation platform working conference "was successfully held in Boao, Hainan A large number of working group managers of the AI medical device innovation cooperation platform gathered here to share the work promotion of their working group, and put forward the outlook for the related work in 2020 Here, the reporter of arterial network sorted out some contents of the meeting, trying to help medical AI practitioners to clarify the direction and progress of the review and approval of the drug administration in the past six months, as well as various problems in the review and approval The overall structure of the innovation platform remains unchanged On the evening of December 27, two new working groups were added The Management Committee of the innovation cooperation platform held a meeting and passed the following resolutions: 1 Revise the standard operation specifications of the innovation cooperation platform for artificial intelligence medical devices; 2 Discuss and pass the measures for the management of member units of the AI medical device innovation cooperation platform and the measures for the management of registration units of the AI medical device innovation cooperation platform; 3 Discuss and pass the establishment plan of the new working group Compared with the organizational structure announced in July, we can see that the AI medical device innovation cooperation platform has added two new working groups, i.e medical data application technology research group and medical AI terminology standardization working group It is worth noting that Zhang Xueli, Deputy Secretary General of the AI medical device innovation cooperation platform, referred to the member unit system and registration unit system She said: "on the premise of being able to comply with the obligations of the platform and the management measures, the institution can sign a cooperation agreement with the equipment audit center to become a member unit The rights and interests enjoyed by member units include members of the promotion management committee; application for establishment of a working group; participation in relevant work according to their own ability; participation in platform related activities; ability to put forward opinions and suggestions on the platform All social organizations registered in the people's Republic of China can register relevant information on the innovation platform to explain their own technology, characteristics, resource advantages and possible work direction The platform Secretariat will review the information and give feedback After the review is passed, they can be registered as the registration unit of the platform In addition, if the registration unit receives the invitation from the working group, it can participate in the corresponding work, or apply to participate in the research project, and obtain the designated public information " It can be seen that whether hospitals, universities, or medical AI enterprises can join in the innovation platform will inevitably have a positive impact on the AI research of the institution itself; the AI medical device innovation cooperation platform also welcomes organizations from all walks of life to participate in the research of medical AI Data security, standard planning What role have many working groups played in the promotion of AI review and approval? In this meeting, 10 working groups including technical regulations, data governance and evaluation database participated in the work report Arterial network selected 7 groups most related to the development of medical artificial intelligence industry for introduction 1 Research on technical regulations and standardization The work of the technical regulation group is mainly to study the policies, regulations and methods of the whole life cycle supervision of the artificial intelligence medical devices and the methods of scientific supervision of the medical devices by using the artificial intelligence technology Peng Liang, deputy director of the first evaluation department of the equipment examination center of the state food and drug administration, said that the basic ideas of the working group include three points, and it is precisely these three ideas that guide the artificial intelligence medical devices Development of equipment review ideas 1、 With regard to the construction of the guiding principle system for the evaluation of artificial intelligence medical devices, the group hopes to build a system to guide enterprises to do well in product application and provide technical support for the development of the whole industry 2、 This paper explores the database building model of distributed AI medical device evaluation database, which is the best model in the current state 3、 Artificial intelligence medical device evaluation database adopts condition evaluation: national ministries and commissions, medical institutions are building relevant databases, but what kind of database can be used as evaluation database? This is a question worthy of consideration Therefore, the working group of technical regulations needs to formulate relevant conditions to evaluate relevant databases and include some databases under construction into the scope of evaluation database system At the same time, these databases need dynamic management to ensure that the relevant conditions of the database can be used for the confirmation of artificial intelligence medical devices At present, the projects such as "review key points of in-depth learning assistant decision-making software" and "independent software for appendix of medical device production quality management specification" in the review guiding principle system have been completed, and the parts such as "human factors and availability" are still under development 2 Construction of Evaluation Database The evaluation database construction working group is mainly led by Shanghai Shenkang hospital development center 301 Hospital, Xiehe Hospital, West China Hospital and other units jointly undertake the promotion work of the working group Combined with the characteristics of artificial intelligence devices, the working group will closely focus on the current data governance, standard system, evaluation technology, clinical evaluation and other hot spots and focus work, and integrate the work of medical institutions Clinical resources and data resources, focusing on the research of key technologies of cross regional, cross institutional and multi-modal evaluation data governance, in-depth research and development of multi hospital, multi heterogeneous and multi-modal evaluation database construction management technology and evaluation database security and open technology, and construction of test database in line with the construction of artificial intelligence medical device innovation cooperation platform In 2019, the work of Shenkang center still focuses on the preparation of AI test database, and other work includes promoting the landing of intelligent image in Shanghai; technological research and development of medical data innovation application platform based on artificial intelligence; building the organizational structure of the specialty Alliance Center, building the specialty research center under the specialty Alliance Center; building the AI regional medical image cooperation and sharing platform And participated in the research and development of the application of artificial intelligence and the demonstration of evaluation technology In the coming 2020, Shenkang center will research and build the "evaluation technology and professional database of artificial intelligence + medical image system" and improve the level of accurate evaluation and quality supervision The specific work will be divided into three parts 1 Improve the system framework of artificial intelligence evaluation database Improve the evaluation database system framework from the standardized collection of evaluation data, data annotation, governance quality control, security sharing and other key technologies to form targeted quality control requirements 2 Carry out the research on the personalized construction management technology of multi hospital area, multi isomerism and multi-modal 3 Collection of the medical artificial intelligence evaluation database, which covers the information of CT pulmonary nodules, fundus, cardiac MRI, coronary CTA, ECG, CT fracture, brain MRI, CT liver and other diseases, and further optimizes the medical test sample data covering 8 types of data, such as electronic medical records, medical images, and electrophysiological signals 3 the real world working group reported that Yu Weihong, chief physician of Peking Union Medical College Hospital, Chinese Academy of Medical Sciences, as a representative, talked about the progress of ophthalmic AI work and the work plan of the real world data application working group in 2020 In 2019, the working group established the "establishment of China's diabetes retinopathy artificial intelligence fundus image standard database", which has completed the collection and sorting of 20000 sugar net fundus images, and is marking them smoothly; three clinical trials of ophthalmic sugar net AI products, including two retrospective studies and one prospective study, were carried out with three enterprises, voxel technology, zhiyuanhui map and Zhizhen Internet The project of "Research on artificial intelligence assisted screening system for main blinding fundus" of Beijing Science and Technology Commission was launched; the "China real world AI research alliance for ophthalmology" was established, and the consensus of fundus color photography data annotation and quality control experts was drawn up In 2020, the working group will comprehensively promote the multi-disciplinary real world research work of each unit in the group; pay special attention to the money management processes such as medical ethics approval, informed consent implementation, data De privacy, signing of relevant agreements on data transfer of cooperative units; pay attention to the clinical trials of AI products, summarize the problems encountered in the experiments and relevant countermeasures, form expert opinions, and submit them to the platform for reference Finally, the consensus draft of medical data annotation and quality control experts in different disciplines was launched 4 The working group of clinical evaluation reported that in December 2019, the principle of knowing the technology of real world data used in clinical evaluation of medical devices began to solicit opinions Next year, on the basis of "key points for evaluation of deep learning decision aid software for medical devices", the team plans to prepare technical guidelines for clinical evaluation of representative products such as diabetic retinopathy deep learning decision aid software for current products with centralized application; and investigate the needs of the industry to carry out scientific research and achievement transformation guided by problems 5 The testing technology research team led by China Academy of information and communication is trying to explore the establishment of a public testing platform for the performance of medical AI At this stage, the technical problems of the industrial evaluation of artificial intelligence medical devices mainly occur in two directions: the technical detection in the process before the product goes on the market and the software update after the product goes on the market In order to solve these problems, relevant projects will be carried out gradually in the past two years Now, the test technology research team has started the work of "medical device software quality test" and "medical device network security test" With these experiences, the test technology research team will complete the preparatory work of "medical AI product performance test" as soon as possible 6 network security Zou Xiaoxiang, deputy director of the second Research Office (New Business Research Office) of the national Internet Emergency Center, delivered a keynote speech related to data security at the meeting He said: "under the guidance of the equipment audit center, relying on the network information office security center, under the guidance of the guiding principles of medical device network security registration technology review (cmde) and the FDA guiding principles of medical device network security management before listing, etc With technical support, we have completed the preparation of network security standards, use case design and test verification of medical X-ray imaging equipment " These work include: 1 Develop 20 network security standard test cases; 2 Carry out security evaluation on 15 products of 8 domestic and foreign mainstream manufacturers; 3 Issue 15 security evaluation reports Therefore, to protect medical data, we must work together from three aspects: end (terminal), management (Network) and cloud (platform) Taking smart medical cloud and gene cloud as examples, this platform not only pushes closed and isolated medical institutions to open and interconnected, but also bears a more severe security test In this regard, Zou Xiaoxiang put forward three suggestions: 1 Study the network security detection technology for biomedical equipment and operable evaluation standards and methods, develop network security standards and specifications, and accelerate the construction of network security evaluation technology system 2 For important systems and key equipment, routine access safety inspection certification and network