Siemens Healthcare Chen Lifeng: the “pain point” and “breakthrough” of medical decision accuracy ccf-gair 2020

On August 7, 2020, the global summit on artificial intelligence and robotics officially opened. Ccf-gair 2020 summit is hosted by the Chinese computer society, jointly hosted by the Chinese University of Hong Kong and Lei, and co sponsored by Pengcheng laboratory and Shenzhen Institute of artificial intelligence and robotics. From the combination of learning and industry in 2016, the industrial landing in 2017, the vertical segmentation in 2018, and the 40th anniversary of artificial intelligence in 2019, the summit has been committed to building the largest, highest specification and most cross-border academic, industrial and investment platform in the field of artificial intelligence and robotics in China. < p > < p > at the medical science and technology special session on August 9, Chen Lifeng, the director of digital medical in China of Siemens Medical, introduced how Siemens Medical used AI to meet the core needs in clinical scenarios. Chen Lifeng said: the most important link in the medical process is decision-making. On average, a medical worker has to make 22 important decisions every hour. These decisions have a direct impact on the quality of examination, diagnosis, treatment and post-treatment management of patients. < p > < p > since 1990, Siemens Medical has applied artificial intelligence to equipment imaging. Siemens Healthcare has long-term experience in the field of AI, and has more than 600 series patents related to machine learning, of which more than 200 are related to deep learning. < / P > < p > for example, in order to solve the problem of missing medical history, Siemens Healthcare provides the solution eHealth, which can provide all the medical history to the doctor for judgment after the patient is authorized. < / P > < p > in the link of appointment inspection and examination, Siemens Medical has also developed the Medicalis system, which can centralize all medical resources on the platform, and then triage patients to corresponding hospitals and departments for diagnosis according to different conditions of patients. This triage scheme will give the optimal path according to the patient’s own region and insurance situation. < p > < p > in addition, Chen Lifeng also introduced AI rad companion, an artificial intelligence film reading platform, which can simultaneously identify six diseases such as pulmonary nodules, emphysema, aorta, heart and spine reconstruction in one CT image, and directly display all diseases in the chest in a quantitative form. < p > < p > Chen Lifeng: first of all, let me introduce Siemens Medical. Siemens Medical has a history of more than 130 years and now has three core business departments: imaging diagnosis, laboratory diagnosis and clinical treatment. < p > < p > in 1999, the year Ali was founded, Siemens Medical has launched the it full platform. In 2014, Siemens Medical also released a cloud based data platform and teamplay, a data cloud platform with its own dosage and dose. Up to now, Siemens has more than 40 AI applications. < / P > < p > next, let’s share the industry trend changes. We believe that the future market dynamics and trend changes will have a great impact on the development of the medical industry. At present, there are some trends and changes in the market that we have never encountered before, such as the aging of the population and the increase of chronic diseases. < p > < p > from the perspective of the large medical market, this trend is the continuous increase of the market. In the short five years from 2015 to 2020, the global major economies’ medical and health expenditure increased from an average of 2.4% to 7.5%. About half of these expenditures, about 4 trillion, were spent on the three major diseases – cancer, cardiovascular and respiratory diseases. < / P > < p > first of all, medical treatment will be more accurate, but the operation mode of precision medicine is very different from the previous operation mode of the hospital, which requires all departments to work together, and even needs us to rethink the transformation mode of diagnosis and treatment. < / P > < p > secondly, improve the patient experience. Now, when you or your family get sick, you will first search the Internet for what diseases you have and how to treat them. This also makes the work of doctors difficult, because you will bring information and research to discuss diseases with doctors. However, from another perspective, it also means that patients pay more attention to and participate in their own medical and health management, realizing the consumerism of medical treatment, and patients are more like consumers. In this process, patients’ cognition and knowledge of the disease are also greatly increased, and they will have more choices. What is decision making? We all make decisions all the time, and medical staff spend most of their time on decision-making every day. These decisions are closely related to everyone’s life and health. < / P > < p > in addition to the clinical aspects, medical workers’ decision-making also involves the operational level. For example, if a stroke patient enters the hospital, if the patient can not be treated in our hospital, he needs to be referred. Where to turn? How? The judgment inside is a typical operation decision-making scenario. < / P > < p > for medical workers, on average, they have to make 22 important decisions every hour. These decisions directly affect the process of diagnosis and treatment of patients, from inspection and detection, diagnosis, treatment, and post-treatment management. However, these decisions often result in a large number of errors and deviations due to the lack of patient history and asymmetric information. For example, when patients arrive at the hospital, they will have a certain understanding of their own disease history, but they do not know the current progress; while doctors can know the current symptoms of patients through testing and testing, but they do not necessarily know all the medical history and medication history, so the final decision will be very difficult. For example, the director of emergency department of a tertiary hospital in Shanghai once introduced a case. The family members of their hospital staff suddenly fainted one day, but they did a series of tests and examinations in the emergency department and found no problems. During the observation period in the hospital, the director and the patient occasionally learned that the patient had mild thalassemia, and immediately thought that thalassemia might cause pulmonary embolism, while the latter would cause pulmonary hypertension and syncope. Finally, it was found that this was the case. < / P > < p > in addition to diagnosis, doctors often need to make a lot of decisions in the process of disposal and examination. For example, the mother of a friend of mine suddenly twisted her back in the process of cleaning the window this year. Although she didn’t pay attention to it at that time, she could not sleep because of the pain, so she went to the top five top three hospitals in China. < / P > < p > after they arrived at the hospital, their most headache was guidance, so they didn’t know which department to go to. So they went to the heart first, and did a lot of CT scan and reconstruction examination in the heart, but they didn’t find any problems. They also went to the orthopedics department, and the orthopedic doctors thought there was no problem, and they inferred that there was something wrong with the nerves. But later they mistakenly went to the Department of Neurology. < / P > < p > in retrospect, about 15% of the imaging diagnoses in these examinations and procedures are wrong or unnecessary or even repetitive. In the process of diagnosis decision-making, if doctors don’t have enough information, it will lead to wrong diagnosis and treatment decision-making or diagnosis and treatment process. However, doctors may not know your past medical history and medication history during the process of prescribing. Such unequal information can make the follow-up process difficult. There are 8 billion cases of diagnosis and treatment in China every year, so it is almost impossible for everyone to follow up. < / P > < p > then, how should we solve these problems of diagnosis and treatment path and process, so as to help patients and doctors make correct decisions in the whole process. I think the best tool to solve these problems is AI. < / P > < p > at present, AI has been fully used in Siemens’ panoramic solutions. There are more than 40 AI applications, and 240000 patients have directly or indirectly experienced the help of AI every hour. < / P > < p > as mentioned just now, doctors need to master the complete medical history of patients to make judgments in the process of treatment, diagnosis and other processes. How to get the complete medical history of patients? < / P > < p > at present, Siemens Healthcare has a set of solutions for medical history, eHealth, which has been launched in Austria. In order to avoid the disclosure of patient privacy, the product will provide all medical history to the doctor for judgment after the patient has authorized the patient remotely or in person through eHealth. < / P > < p > Siemens Medical also has a system called Medicalis, which has been launched in the United States. The system can concentrate all medical resources on the platform, and then according to the different conditions of patients, patients can be triaged to corresponding hospitals and departments for diagnosis. In the aspect of < / company >. We hope that this product can be compatible with multiple manufacturers, provide multi organ comprehensive diagnosis, and finally automatically generate display services.

for example novel coronavirus pneumonia novel coronavirus pneumonia, we found that SIEMENS CT device scanning images, using AI-Rad Companion for new crown pneumonia positive and negative judgment, will be more accurate than other manufacturers’ products. < / P > < p > the reason behind this phenomenon is that the products of other manufacturers will have some false negative and false positive results in the early judgment of pneumonia due to the convolution kernel problem. Siemens Medical’s AI rad companion has been developed in a Supercomputing Center in Princeton for many years, and has reached the clinical stage in China. It has cooperated with many hospitals in research and development. Pulmonary nodules and pulmonary nodules can not only be seen in the form of pulmonary tuberculosis and pulmonary tuberculosis, but also be seen directly in the form of pulmonary nodules and pulmonary nodules. < / P > < p > for diagnosis and treatment decision-making and treatment, all methods are also in continuous innovation. This is just like navigation. In the past, everyone needed to buy a National Atlas. After GPS came out, some roads were upgraded, but the software was not upgraded, which led to a wrong way. But now the mobile phone can complete all the navigation problems. < / P > < p > the diagnosis and treatment path of patients is the same. In the past, a whole body skeleton was placed on the edge of many clinics, but now it is no longer needed. At present, the decision-making of diagnosis and treatment needs only two things. First, we need to know the various test indicators, and then we need to know the comprehensive medical history and the state of the patient. < / P > < p > therefore, many manufacturers have proposed the concept of digital Twin / digital twin. They digitize all medical indicators of this person, including laboratory, medication, image data, gene, etc., and then integrate these data with literature, guidelines and other knowledge bases to select the best treatment plan for patients. < p > < p > AI rad companion also has a follow-up module, which can use a small household device to manage the changes of various indicators of patients. In this way, the above-mentioned 8 billion person time follow-up problem can be solved. With the help of artificial intelligence, the equipment can automatically put forward early warning and other intervention plans. In this way, a nurse can even follow up thousands of people. < / P > < p > we have already introduced the Siemens patient diagnosis and treatment pathway and digital solutions. Let’s go back to understand the driving forces of artificial intelligence. I think it mainly includes three aspects: < / P > < p > so around these three aspects, AI also plays several roles: first, it improves efficiency and productivity; secondly, it provides clinical decision support; finally, it provides hospital with Patient management. < / P > < p > from a purely technical point of view, the biggest problem in AI application is the data island problem between various departments and institutions, which may lead to potential problems