AIGC wave drives medical revolution, how to build artificial intelligence-assisted medical examination?

21st Century Business Herald reporter Tang Weike reported from Guangzhou

The success of ChatGPT has set off a wave of generative large-language model development. With the successive launch of general large models, the domestic "100-model war" has risen rapidly. Now, this war has begun to sweep into vertical fields, and large models based on medical, financial, education and other applications have begun to appear.

The emergence of AIGC technology has made the intelligent development of the medical examination industry also full of imagination. At present, medical testing services are becoming professional, precise and personalized. Whether it is the selection of projects or the interpretation of results, it is necessary to combine a large number of cases, papers, doctors' experience and other dimensions of analysis. AIGC has the ability to answer questions, generate, summarize and dialogue. Around the mutual transformation of text and images, it can be used for text generation, knowledge Q&A, data enhancement, scientific research, talent training, etc.

Once AIGC technology is truly implemented in the field of medical examination, it is expected to help doctors rapidly improve their clinical level and improve the efficiency of medical resource utilization. At the same time, it also assists doctors to quickly complete disease diagnosis, improve the accuracy and reliability of diagnosis, and allow patients to enjoy faster and more accurate diagnosis and treatment services.

Fan Xiaojuan, a partner at Zhong Lun Law Firm, told the 21st Century Business Herald reporter that AIGC's technological change does urge all walks of life to think in their own fields, and what far-reaching impact such a large model era will have on their professional fields. For the medical testing industry, AIGC has become an important force in building a medical examination ecosystem.

Open space for medical applications

From the perspective of market demand, when ChatGPT appears, large models will reshape many industries, and large models will become the operating system of artificial intelligence and enter thousands of households. From chips, management and loans, large models will bring more and more new research directions in hardware and network security, and the application of AIGC will be more extensive.

AIGC has a broad space for development in the medical examination industry, and through big data models, creating intelligent general practitioners can help the medical examination industry.

Recently, the 2023 "Domain See Cup" Medical Inspection AI Developer Competition, guided by Guangzhou Science and Technology Bureau and Guangzhou Municipal Bureau of Industry and Information Technology, hosted by Guangzhou Jinyu Medical Laboratory Group and Guangdong Artificial Intelligence Industry Association, co-organized by Guangzhou Artificial Intelligence Public Computing Power Center, and supported by Huawei Cloud Computing Technology Co., Ltd. was also officially launched.

Zhang Chao, founder and CEO of Doctor Zuo, told the 21st Century Business Herald reporter at the launching ceremony that ChatGPT learns wisdom and summarizes through data. True medical knowledge is decentralized in the hands of a few, and ChatGPT is able to fuse top-notch, leading, innovative knowledge against the moderate majority.

Through continuous popularization analysis and expert feedback on a wide variety of data, the generation of valuable data models, plus some prefabricated decision-making knowledge, the future large model can achieve a very good effect.

Virtual doctors built on large models can help in healthcare. Liu Shanrong, director of the Department of Experimental Diagnosis and doctoral supervisor of Shanghai Changhai Hospital, said that if one or several regional test GPTs can be built through ChatGPT in the future, and through the collection, learning and integration of high-end doctors in large third-class hospitals for certain diseases, many third-class hospitals can appear, which can also be called regional ChatGPT.

Development still has problems to overcome

In the medical field, it is very difficult to produce AIGC. "AIGC is mainly used in media, e-commerce and other fields. These fields do not require standard answers, while in the medical field, production content must be precise. Chen Junlong, dean of the School of Computer Science and Engineering of South China University of Technology and a foreign academician of the European Academy of Sciences, told the 21st Century Business Herald reporter that in the field of medical examination industry, the front-end and back-end should add a decision-making model, and a model that generates content cannot make an accurate diagnostic model. The future should be a combination of production models plus professional decision-making models.

Zhang Chao also believes that the construction of decision-making models is very important: "Human communication and speech are expressed word by word, and the essence is that the brain is thinking. The artificial intelligence model needs to be guided by a more accurate abstract logic to express the content, which is both a decision-making model. ”

The medical field is particularly demanding on data. Liu Shanrong said that the data in the medical examination field has the problem of data silos, and if it cannot break through the bias of each hospital on patients, then the model must be biased. And the degree of data specialization is high, the degree of dependence of medical examination on experts is also relatively high, to tell artificial intelligence through expert judgment, structuring and standardization is to test artificial intelligence to do.

Difficult diseases are complex and diverse, and each disease needs to build artificial intelligence models, for different diseases, establishing accurate artificial intelligence models requires a huge amount of work, and the establishment of test models needs to break through the original solidification. Professor Liu Shanrong said, "When building an inspection model, no matter how accurate, it is a solidified model. But clinicians are not. Clinicians will relay to patients what the probability of this disease is, and AI models have not been able to do so far. AI models need to establish a process of self-assessment. ”

AIGC applications need to meet legal compliance requirements. After the big model is generated, the risk comes. To break through the application prospects of AIGC in the medical examination industry, it is necessary to meet the requirements of compliance. Fan Xiaojuan said that the risk of compliance is composed of several aspects: "First, the training data may have compliance risks, intellectual property and privacy data, and whether the source of the training data is legal; The second is that the generated content is risky, and the risk of generating directly bad content is generated; There are also harmful content that is not in line with socialist values; The third is privacy leakage, network security and data security, OPERN Ai also has data leakage; The fourth is criminal risk. ”

As data generated by large models, personal data needs to be protected. There are special requirements for medical data, and some important sensitive data needs to comply with personal information protection regulations. In addition, when large models and overseas cooperation, the authentication check of data is particularly important, and sensitive information involving 10,000 people needs to be evaluated for security.

Intelligent transformation

By building an industry model, the level of intelligence in the medical field can be improved, and the transformation of the medical examination industry from digital to intelligent can be promoted.

Professor Tian Qi, Chief Scientist of HUAWEI CLOUD Artificial Intelligence and Academician of the International Eurasian Academy of Sciences, and IEEE/CAAI Fellow, said that medical institutions are not good at building general models and should leave it to other enterprises to build basic models. On the basis of the general model, the industry model can be built by adding the amount of industry data. Industry models are what medical institutions are good at, and they have rich resource data to establish a sound industry model.

The sharing of data and computing power helps to build accurate industry models. Professor Tian Qi mentioned, "Each industry has the advantages of each industry, you can establish a data alliance for enterprise cooperation, achieve the sharing of specific data through a certain mechanism, or the entire medical industry can join forces with some other enterprises to build a computing power alliance." ”

The application of AIGC plays an important role in the whole chain of the medical industry, from the transformation of medical and pharmaceutical regulatory agencies from digitalization to intelligence, intelligent applications promote the development of the whole chain.

The sinking of high-quality medical resources is the development direction emphasized by the state in recent years. AIGC is applied to the medical examination industry to build intelligent medical examination, which is of far-reaching significance to promote the sinking of high-quality medical examination resources. Professor Liu Shanrong said, "The national level emphasizes the sinking of high-quality medical resources, and the use of artificial intelligence to assist medical examinations is the easiest way to achieve the sinking of high-quality medical resources. It can help remote or grassroots hospitals to achieve early detection and treatment of complex diseases and major intractable diseases. The realization of intelligent auxiliary medical examination is of great significance to the development of medical industry. ”

(Intern Huang Yuanxuan also contributed to this article)

For more information, please download 21 Finance APP