On September 8, Tencent Health announced a medical model, with a new multi-scenario AI product matrix such as intelligent Q & A, family doctor assistant, and digital intelligence medical imaging platform. The drug discovery platform "Cloud Deep" (iDrug) revealed a new protein structure prediction algorithm framework tFold.
Wu Wenda, president of Tencent Health, pointed out that the digital intelligence upgrade of medical health has changed from "first try" to "full embrace", and the application of artificial intelligence and various digital intelligence tools will accelerate the popularization; Tencent will work with partners to build a "TRUST" model, strengthen technological innovation and digital infrastructure, and promote the universal popularization of "digital intelligence medical care".
Tencent Health President Wu Wenda: Fully embrace the digital intelligence upgrade of medical health
Let the big model understand medicine better
Zheng Yefeng, an outstanding scientist at Tencent and director of Tencent Tianyan Lab, announced the development process of the medical model. Tencent’s newly released full-link self-developed hybrid model has a scale of over 100 billion parameters and over 20,000 billion tokens of pre-training corpus, with strong Chinese creation ability, logical reasoning ability in complex contexts and reliable task execution ability. With this as a base model, continue to add medical Knowledge Graph and medical literature covering 2.85 million medical entities, 12.50 million medical relationships, and covering 98% of medical knowledge, so that the big model can further master professional medical knowledge. After 30 million multi-tasking fine-tuning of Q & A dialogues covering patients, doctors, pharmaceutical companies and other scenarios and medical processes, as well as reinforcement learning of data labeled by 360,000 group of expert doctors, Tencent’s medical model is more professional and accurate in handling medical needs, while also taking into account patient care, closer to human doctors.
Tencent outstanding scientist, Tencent Tianyan laboratory director Zheng Yefeng announced the medical model R & D process
At present, Tencent’s medical model includes scenario models such as copy generation, intelligent Q & A, medical record structuring and retrieval, image reporting, and auxiliary diagnosis, which can be embedded in the whole process of medical treatment, and can achieve comprehensive improvement in medical service levels and quality in application scenarios such as department guidance, doctor recommendation, pre-consultation, doctor-patient dialogue, automatic medical record generation, and intelligent hospital customer service.
Intelligent upgrade of the whole process of medical treatment
Wu Zhigang, general manager of Tencent Health User Platform, introduced that relying on the medical Knowledge Graph and medical model, AI guided and assisted medical applications have been fully upgraded, including medical intelligent question and answer, digital human medical assistant, intelligent free consultation, automatic medical record generation, AI rational drug use, intelligent follow-up management, patient whole process management and other applications, to provide more accurate assistance for the whole process of medical decision-making, and to help patients experience medical treatment and improve the efficiency and quality of clinicians and pharmacists.
In addition to medical intelligent Q & A, the application in the pre-diagnosis stage has also upgraded the digital human medical assistant, which can provide 7 * 24h intelligent customer service and professional popular science services; in the diagnosis center section, the ability to learn million-level doctor-patient dialogue and more than 3,000 disease deduction and analysis based on large models has been upgraded. Applications such as auxiliary consultation with stronger simulation of free dialogue authenticity, automatic generation of medical records that more conforms to writing specifications, and aided diagnosis an treatment that can give interpretable basis have been upgraded; in the post-diagnosis stage, the new model of artificial intelligence + Internet medical treatment has upgraded intelligent follow-up management and patient whole-process management.
At present, Shanghai Renji Hospital has taken the lead in using the Tencent medical model, focusing on Internet hospital senior-friendly services, to create a highly personified virtual digital image "Xiaowei Nurse" to provide warmer and more convenient medical services for elderly patients.
Multi-scenario driving the popularity of AI applications
In terms of AI-driven drug discovery, Tencent’s "iDrug" platform has the ability to accelerate the discovery of both small and large molecule drugs. According to Liu Wei, head of Tencent AIDD technology, in the field of protein structure prediction, the "cloud deep" (iDrug) platform has developed a new algorithm framework tFold, which has been proven by international protein structure prediction and evaluation platforms for many times; in the prediction of ADMET properties of drugs, 70 + ADMET properties have been developed and launched, and it can surpass mainstream software through data training such as combining physical characteristics and drug company cooperation. Two skeleton transition molecular generation algorithms have been developed to find nM-level lead compounds, which have been effectively proven in 3-4 projects. At the same time, reinforcement learning technology has been introduced to achieve 97% of the generated molecules in the generation of drug small molecules.
In terms of supporting grassroots service capacity enhancement, Tencent Health also released a new family doctor assistant. Wang Qianyi, general manager of Tencent Health Strategy and Medical Products, pointed out that the use of WeChat and WeCom to connect the communication channels between doctors and patients, as well as the help of medical models, will provide doctors with more intelligent signing, consultation, follow-up, and education capabilities. Based on the capabilities of medical models, family doctor assistants can be transformed into "grassroots AI know-it-all" to provide residents with fine health management.
In terms of medical imaging research and application, Tencent Miying has fully upgraded its digital medical imaging platform, focusing on digital medical imaging cloud applications, scientific research and management. At the same time, it is equipped with artificial intelligence analysis capabilities to provide integrated medical imaging solutions for production, learning, research and management. It provides comprehensive support in the field of medical imaging for medical institutions, scientific research institutions and science and technology innovation enterprises. The platform will also further open more than 20 AI engines developed by Tencent to help scientific research institutions independently develop and train AI models.
Constructing the "TRUST" Model of Digital Intelligence Healthcare
At present, Tencent Health Mini Program has connected more than 3,000 registered public hospitals, more than 5,000 physical examination institutions, and vaccine services in more than 200 cities; medical insurance electronic vouchers cover more than 800,000 institutions and more than 20,000 hospitals; WeChat channel electronic health card users exceed 600 million, access to more than 5,000 hospitals. Tencent Health further strengthens the connection between users and medical services.
Wu Wenda, president of Tencent Health, pointed out that in terms of technological innovation, Tencent Health continues to invest in medical AI-related technology research, accumulating more than 1,000 patents, and accelerating the transformation of cutting-edge technologies into medical scenarios; at the same time, Tencent’s accumulation of digital infrastructure such as computing power supply, data storage, and security protection will provide a reliable underlying guarantee for the medical industry; in addition, with the help of WeChat, Tencent conferences and other extensive user capabilities, artificial intelligence applications can widely help medical inclusion; to this end, Tencent Health will further promote its own capabilities, technology and product openness, and build a digital medical standard consistent with the industry to create a replicable innovation benchmark.