High-quality medical resources in China are lacking and unevenly distributed. This means that for the more than 4 million new tumor patients each year, a large number of patients find it difficult to access medical services that are suitable for them. Uneven distribution of medical resources, even in the first-tier cities have to face the problem of medical institutions are difficult to provide continuous, high standard of medical services - how to solve the shortage of diagnostic and treatment resources, as well as how to achieve as much as possible the expansion of diagnosis and treatment standardization coverage throughout the country? These are all issues that need to be addressed in the field of diagnosis and treatment.
The development of technology provides us with the opportunity to solve this pain point. With the continuous maturation of artificial intelligence, the industry is exploring the ability to achieve replicable services through artificial intelligence technology to allow machines to achieve standardization of services for each patient, and to reduce the cost of services to a sufficiently low scale of services. This ****ty problem is gradually being solved.
CSCO AI is one such solution. It is jointly developed by the Chinese Society of Clinical Oncology (CSCO) and the national high-tech enterprise Zhejiang Haixinzhihui Technology Co. It combines multi-dimensional knowledge in specialized fields such as CSCO clinical diagnosis and treatment guidelines, experts' clinical practice experience, high-level clinical evidence, adverse reaction management system, etc., to assist clinicians in formulating more standardized and precise treatment plans. At present, Maxim's has successfully built a leading new platform for intelligent tumor services in China, providing patients with one-stop management services for the whole course of their illness, including diagnosis, treatment and rehabilitation. This is the cornerstone of the industry-leading Knowledge Graph capability built by Maxthon.
Knowledge graph is a concept of knowledge network system proposed by Google in 2012, which simply means that scattered information is connected through semantic relationships and transformed into a visual knowledge network. Knowledge graph technology can model, organize, and manage medical data in a unified way, which not only effectively describes and mines the relationships between medical knowledge, but also provides strong support for higher-level medical applications such as assisting clinical diagnosis and treatment decisions, and medical Q&A.
With a knowledge graph as a way to express knowledge, how to utilize it becomes a question of choice in front of HMI. Even if we focus on the medical field, to build a general medical knowledge graph, it can only be the knowledge of some diseases, the definition of combing and integration. "General medical knowledge mapping in the clinic should have great limitations, it is only suitable to do some simple popularization, diagnostic guidance and initial judgment, if you want to go deeper into the disease and clinical diagnosis process, and even the whole disease management, tracking, follow-up details to go, will encounter a variety of problems: the differences between the disease itself and the entire clinical treatment of the relevant knowledge systems The differences between each disease and the entire clinical treatment-related knowledge system extend, are interrelated and affect each other, making the generalized knowledge mapping system in the choice of drugs, post-rehabilitation and other aspects will encounter difficulties," said Eddie Li, founder of Haixin Zhihui.
Therefore, Haixin Zhihui chose to use the "deep and detailed" approach combined with artificial intelligence to build a knowledge map of the entire oncology service system, which will clearly define all the knowledge related to oncology, and this is undoubtedly the most perfect application of the knowledge map direction.
"We spent several years in the early stage to define the knowledge of the tumor from the beginning of the diagnosis to the tumor into the treatment, and then to the patient's whole course management, which is to build up a more comprehensive tumor knowledge graph. The reason why we chose the field of tumor is that this disease is characterized by a more complex diagnostic and treatment system compared to other diseases, and the time span of the continuous treatment process is much longer, so the knowledge graph in the field of tumor treatment can maximize its role," explains Eddie Li on why he chose to apply the knowledge graph to the tumor track.
The second reason is that knowledge of tumor diagnosis and treatment is constantly being updated every year. With the advancement of technology, certain diseases have been completely researched, and there is no fundamental difference in treatment strategies between ten years ago and ten years from now. In contrast, humans still do not fully understand the pathogenesis and mechanisms of tumors, and their treatment is an ever-changing process. Globally, the level of standardization of oncology treatment in healthcare institutions is not high. Medical professionals are facing a very big challenge - how can they keep up with the cutting-edge academic developments at home and abroad, and be able to accurately grasp the latest advances in clinical research and the expansion of therapeutic modalities, accurately grasp the latest drugs and therapeutic modalities, and apply them to the most suitable patients? "If a city hasn't built a new road in 30 years, then people living in that city don't really need maps and navigation - because nothing has changed. However, if the city is growing at a rapid pace, with 'three years of small changes and five years of big changes', then even people who have lived in the city since they were young might need navigation to guide them when they are traveling", said Eddie Ying Li, who used a simple example to illustrate the ever-changing nature of oncology treatments.
Anti-tumor systemic therapies have changed dramatically in the last decade. Ten years ago, chemotherapy might have been the mainstay, but today, there are more clinical options: targeted drug therapy, immunotherapy, to the latest CAR-T cell therapy, and the list goes on. This also shows that the country has been working hard to try to establish a more modern and standardized diagnosis and treatment service system, which can allow tumor patients to get the most appropriate, standardized, standardized and suitable treatment.
The assisted decision-making system is the best tool to help doctors standardize treatment, and CSCO AI's intelligent assisted decision-making system automatically generates diagnosis and treatment recommendation reports and submits them to high-level experts for review and feedback after patients upload various medical diagnostic information through the APP, and the AI model continues to receive closed-loop training based on the results of the experts' review. Behind this set of operational systems is the combination of the oncology knowledge graph and neural support decision-making algorithms that Haishen Zhihui has built with the strength of its experts.
Through the comprehensive knowledge system of the knowledge graph, CSCO AI not only realizes the standardization and homogenization of treatment in intelligent assisted diagnosis and treatment, but also realizes the whole management of out-of-hospital patients, so that multiple **** win - hospitals improve the overall survival rate of the treatment, patients improve the time of survival, quality of life and treatment adherence, pharmaceutical companies also improve the survival time, quality of life, and the quality of life, and pharmaceutical companies also improve the quality of life. treatment adherence, pharmaceutical companies have also improved their ability to support patients through the combination of data-enabled and digital platforms, realizing accurate diagnosis and treatment and precise rehabilitation support. At the same time, the clinical new drug service system based on the management of the whole disease process provides patients with precise matching of clinical research, providing them with a greater choice of possibilities.
Relying on the precise condition, establish the content service capability for the whole treatment cycle of patients, and realize the high-value community platform by multi-dimensional means such as the establishment of medication mind, positive experience incentive, and strong service at key nodes. These are the features of the Haixin Zhihui service.
Knowledge mapping is the basis for assisting diagnostic and treatment decisions, and high-quality data and specialized knowledge systems are the cornerstones of knowledge mapping, so how to build the data layer and knowledge system becomes the key to the quality of knowledge mapping.
CSCO AI's knowledge graph is not created out of thin air, but rather, it is an effective computer-structured deposition of the knowledge that is currently being used by clinical experts to facilitate re-use. The most important aspect of this process is to identify the knowledge base that influences clinical decisions. Only after this knowledge is clearly defined can a model be designed to address it.
Oncology is an evidence-based medicine that must be supported by sufficient evidence to influence clinical decision-making behavior. However, tumors are diverse and specific - each with at least 3,000 pieces of high-level clinical evidence for clinical treatment decisions. It is only by combining these evidences that the tumor-related diagnosis and treatment system and knowledge system can be basically constructed. "What kind of population, characteristics, molecular typing, gene loci, and what kind of treatment under what circumstances can achieve better efficacy, this is the first level of foundation", said Eddie Lee.
After establishing the first level of foundation, the next step is to consider the fit between the clinical treatment plan and the patient's constitution, such as whether the patient's body can withstand it, and whether his or her underlying diseases and pre-existing comorbidities will have an impact on the choice of treatment. This is the second level of performance, the expanded body of knowledge in clinical utilization.
In this process, HMI CSCO AI is building the lowest level of core knowledge of a single tumor with the entire treatment knowledge of the condition at its core, and then further expanding the construction of the knowledge system in clinical application. This ensures that there is core evidence of professionalism, but also reveals the professionalism of medical treatment in the process of clinical application.
What is less well known is that "communication" is the biggest challenge in the knowledge graph building process. To extract knowledge and relationships from big data, the knowledge graph requires the coordination of different professionals. Therefore, the use of knowledge graph technology in specific business areas is actually a cross-border behavior. The difficulty lies in who is going to integrate the cross-border knowledge, which is the first difficulty. This means that the builder of the oncology knowledge graph needs to master the engineering algorithms and understand all the definitions of terms and basic medical knowledge of oncology. "There is a natural barrier, called the knowledge barrier, to how to apply this knowledge graph technology to a new business area. Knowledge mapping technology is mastered by the science and technology department, mastering medical knowledge is the medical department, both have their own mindset, it is difficult to carry out a professional disciplinary dialogue," said Eddie Lee.
The second barrier is the understanding of professional knowledge and grooming. Eddie Lee believes that only after the systematization of knowledge combing can be carried out after the initial construction of the knowledge map. After that, the business modeling on top of this map; in the business application, there are new professional clinical knowledge input. The third barrier is how the computer applies this new knowledge, and how to put the new knowledge on a specific patient case for effective reasoning and decision-making.
Therefore, the biggest difficulty in building a medical knowledge graph is interdisciplinary integration, and the difficulty in integration is that the entire process is more demanding for both disciplines. This is precisely where the advantage of Haixin Zhihui lies - it is the AI strategic partner of CSCO, the Chinese Society of Clinical Oncology. Promoted and assisted by the Society, domestic head oncologists have spent a lot of time and energy to help Maxim's technicians understand the clinical paths and clarify the system based on a high degree of awareness of their social responsibility. CSCO AI is the first intelligent assisted decision-making product developed based on China's diagnostic guidelines and diagnostic practices, which is based on China's national conditions and therefore more suitable for China's diagnostic and treatment application scenarios.
Not only that, CSCO AI can also realize its higher clinical value in multiple scenarios. For example, the interconnection of upper and lower level hospitals - patients consult online through the APP, and doctors use CSCO AI as a tool to link upper and lower level hospitals, leading to standardized diagnosis and treatment in different levels of hospitals.
CSCO AI can also be a research tool, through real cases to carry out clinical research, from a variety of perspectives to explore the intelligent decision-making system on patient treatment and clinical application of help. In addition, the treatment plan recommendations provided by CSCO AI follow evidence-based medicine, and can be used as a quality control tool for standardized treatment in departments, hospitals, healthcare consortia, and governments to improve the comprehensive level of regional standardized treatment. Teaching hospitals can also use CSCO AI as a case analysis learning and assessment tool for residents. As a reference and management tool for tumor patients' treatment decisions, CSCO AI can also be used simultaneously in multiple scenarios combining MDT discussions, physician learning, room visits, case discussions, and so on, so as to enhance the overall clinical outcomes.
The underlying layer of CSCO AI's tumor knowledge graph, although complex, is excellent in terms of ease of use. For patients, all they need to do is follow the prompted steps given by the platform, with no difficulty at all in getting started. For doctors, this AI-driven full-course management system from Haixin Zhihui realizes system tracking capability, monitoring capability, follow-up capability, and data management capability in the process of patients' long-term treatment management, and ultimately achieves a balance of efficiency among doctors, nurses, and patients.
Another highlight of this knowledge graph-based case management service is the introduction of the case manager role, which bridges the gap between physicians and patients. The case manager not only assists the department in daily patient management, but also follows up on patients' treatment and rehabilitation, and monitors and assists the doctor in dealing with out-of-hospital adverse events. With the case manager's follow-up, the patient's treatment and management is also extended from the hospital to outside the hospital, and the treatment behavior is continuous.
In this regard, Eddie Li believes: "The field of oncology diagnosis and treatment will definitely enter a patient service-centered era in the future. Through the help of artificial intelligence, multiple roles can work together efficiently to help patients complete the entire treatment management service. Of course, there is no way for this system to be accomplished overnight at present, and it needs to evolve in actual use."
Although the construction of the underlying so complex knowledge map, but for the knowledge map and the ability to assist the diagnosis and treatment system boundaries, Haixin Zhihui is a clear perception. At present, whether it is in the field of pathology, medical imaging, diagnosis and treatment based on AI technology to build models, quantitative analysis, feature correlation, decision-making recommendations, efficacy prediction and other aspects of the attempt, are only to help doctors improve diagnosis and treatment efficiency, accuracy and the ability to predict the efficacy of the empowered medical behavior, and will not interfere with the doctor's right to judgment and decision-making.
The technology in healthcare needs to be treated with respect.