Nowadays, artificial intelligence is no longer a proper term in science fiction. It has broken through the technological turning point from "unusable and unusable" to "usable" and has entered a period of explosive growth. In the medical field, artificial intelligence can quickly diagnose diseases, perform operations, and carry out health monitoring. In April 2018, the General Office of the State Council issued the "Opinions on Promoting the Development of "Internet + Medical Health"", which clearly proposed the promotion of "Internet +" artificial intelligence application services. This means that "artificial intelligence + medical care" will substantially change people's medical treatment model and help build a healthy China.
--editor
"Medical skills" surpass young doctors
The "artificial intelligence doctors" who have undergone continuous training have increasingly developed their "eyes", "ears" and "brains", which involve more and more diseases and wider fields.
Can "see" images, "read" medical records, "perform" operations, "do" examinations, and give clinical diagnosis recommendations; "medical skills" surpass young doctors, and are comparable to senior doctors in some fields. It is the "artificial intelligence doctor".
Like human doctors, "artificial intelligence doctors" also see a doctor by means of seeing, smelling, and hearing.
Take lung nodules as an example. For lesions as small as 1 mm, the reader needs to look at CT images one by one to find out the size and density. Senior doctors read one film in 10 minutes on average. Large hospitals have more than 100,000 films per day. The work of doctors is intense and heavy. Nowadays, some hospitals have begun to introduce artificial intelligence system for screening, and the reading time has dropped to one and a half minutes.
"Artificial intelligence doctors" are not only highly efficient, but also more sophisticated and comprehensive in diagnosis. In Union Hospital Affiliated to Tongji Medical College of Huazhong University of Science and Technology, the artificial intelligence system developed by Yitu Medical can not only detect lung nodules, but also describe the characteristics of the lesions in multiple dimensions, including size, volume, density, CT value, and nodule characterization It can cover 6 common benign and malignant signs-lobes, burrs, pleural depression, cavities, vacuoles, and calcification. The system developed by Ali Health screens peripheral diseases together, including alveoli, arteriosclerosis, lymphatic banding, increased lung density, and cords.
Artificial intelligence is equipped with "eyes" that can read standardized images and screen out lesions. The Union Hospital of Tongji Medical College of Huazhong University of Science and Technology tested the "artificial intelligence doctor" clinically and found that the detection rate was 95.78%, but the false alarm rate was only 2.63%. In 2018, 60 imaging doctors in the hospital interpreted more than 150,000 imaging cases through the AI system.
In addition, "artificial intelligence doctors" can also check esophageal cancer, diabetic retinopathy, colorectal tumors, breast cancer and other diseases, and even check the bone age of children, with a technical level no less than that of senior doctors.
Artificial intelligence also has sensitive "ears." In Luyang District, Hefei City, Anhui Province, the University of Science and Technology iFlytek Intelligent Medical Assistant has been employed in a community health service institution in 2018. During the doctor-patient communication process, the intelligent medical assistant generates and automatically extracts medical records through big data and intelligent voice technology. You can also query similar cases, clinical guidelines and symptomatic drugs. At present, the system has completed more than 7000 auxiliary diagnosis suggestions.
Recently, "artificial intelligence doctors" have also been equipped with "brains." At the Guangzhou Women and Children's Center, the artificial intelligence system learns to "understand" the medical records, and then gives a diagnosis like a human doctor. The doctor enters the patient's main complaint, symptoms, personal disease history, examination results, imaging examination results, medication use and other information into the medical record text, and the system automatically converts the free medical record text into standardized, standardized and structured data. After the artificial intelligence system "understands" the medical record, it then gives the diagnosis result.
Does the "artificial intelligence doctor" have a high diagnostic accuracy? Taking respiratory diseases as an example, the accuracy of the artificial intelligence for the diagnosis of upper and lower respiratory diseases is 89% and 87%, respectively, and the diagnosis accuracy for different types of asthma is between 83% and 97%.
"Artificial intelligence doctors" who have undergone continuous training, "eyes", "ears" and "brains" are becoming more and more developed, covering more and more types of diseases and wider fields, including clinical assistants, auxiliary diagnosis and treatment, medical imaging, genetic testing, health Management etc.
Artificial intelligence relies on massive data
The degree of standardization of data in various disciplines affects the degree of application of artificial intelligence. Different hospitals have different equipment and different data dimensions
The ability of doctors depends on the study of medical colleges and the accumulation of clinical experience. What does "artificial intelligence doctor" rely on? Rely on massive data and cloud computing capabilities. After "eating" the data, after continuous training in clinical thinking, the system can see a doctor like a human doctor.
After "eating" massive amounts of data, the machine can not only be a doctor, but also can do scientific research, teaching, management, etc., helping doctors and hospitals improve their scientific research and improve their diagnosis and treatment capabilities.
In West China Hospital of Sichuan University, Yitu Medical included full-dimensional desensitization clinical data of lung cancer patients admitted to the hospital from 2009 to the present, and opened up multiple system data such as clinical outpatient, hospitalization, medical records, pathology, etc., and established the first domestic lung cancer clinical research Intelligent disease library. With this disease library, the research capabilities of the hospital's many departments related to lung cancer diagnosis and treatment have been greatly improved, and other medical consortia have also benefited a lot.
Alibaba Health's artificial intelligence medical care has been upgraded to version 2.0. In addition to clinical, it also has text research and imaging research platform functions, providing virtual patients and VR simulation surgery for teaching.
In the clinic in Renzhuang Village, Jiaxian County, Henan Province, the reporter saw the WeChat artificial intelligence auxiliary diagnosis system-general practice auxiliary diagnosis and treatment system, Xuanhutai Chinese medicine intelligent diagnosis and treatment system. Village doctor Zhang Qiaofen simply entered the patient's basic symptoms, medical history, etc., and immediately saw the possible prompts for related critical illnesses and common diseases. "We seldom come into contact with critical illnesses, but we are still worried that if we are misdiagnosed, the villagers' treatment will be delayed."
According to reports, this general practice auxiliary diagnosis and treatment system has studied more than 5 million documents, tens of millions of medical records and health files, and currently covers more than 2,000 diseases and more than 5,000 symptoms, with a hit rate of 90%. The Xuanhutai Chinese Medicine Intelligent Diagnosis and Treatment System has accumulated more than 2 million auxiliary prescriptions.
"Grass-roots doctors have weak service capabilities, and artificial intelligence-assisted diagnosis and treatment can make up for the lack of resources and improve the service level of doctors." Chen Qiulin, director of the Social Security Research Office, Institute of Population and Labor Economics, Chinese Academy of Social Sciences, believes that medical artificial intelligence can improve medical diagnosis The degree of accuracy can also replace the operations in some high-precision surgeries, and can also replace part of the human resources in some medical services, thereby reducing medical expenses.
The data for artificial intelligence learning comes from the clinic and has to be converted into a structured format, and then a model is made to train and learn according to the clinical diagnosis and treatment thinking, and calculate the result. Data is the key. The degree of standardization of data in various disciplines affects the application of artificial intelligence.
Ni Hao, President of Yitu Medical, told reporters that non-standard medical data is a common problem. Although the images are a batch of well-standardized data, there are still big differences between different hospitals. Each hospital has different equipment and different data dimensions. High-quality data is very rare, and it takes more algorithms to structure the data before it can be used.
In 2018, Ning Guang, an academician of the Chinese Academy of Engineering and deputy dean of Ruijin Hospital affiliated to Shanghai Jiaotong University School of Medicine, led the team and Ali Health Artificial Intelligence Laboratory to jointly develop the "Ruining Sugar" artificial intelligence doctor. In the process of advancement, Ning Guang also discovered data problems, such as lack of standard data, inconsistent disease diagnosis standards, follow-up data scattered in various hospitals, and poor data universality.
The degree of data standardization is related to the maturity of the subject and the external conditions required for diagnosis. For example, the imaging field has been a unified standard from the beginning, and the degree of digital development is relatively high; dermatological diagnosis is more dependent on pictures and videos to identify lesions, etc., and artificial intelligence in these disciplines has developed rapidly.
The machine and the doctor cooperate to see a doctor
Medical care is not only about diagnosis and treatment, but also involves the interaction between doctors and patients. In particular, doctors’ comfort to patients is irreplaceable.
Will artificial intelligence medicine replace doctors? To be sure, not yet.
In 2017, the State Council's new-generation artificial intelligence plan proposed to "Develop human-machine collaborative surgical robots and intelligent diagnosis and treatment assistants" and "Develop man-machine collaborative clinical intelligent diagnosis and treatment solutions." This means that artificial intelligence is just a doctor's assistant.
Some artificial intelligence researchers have proposed that only by understanding the psychology and clinical thinking of doctors and letting artificial intelligence learn this kind of thinking can the real medical artificial intelligence be. However, this difficulty seems difficult to break through.
"I have a certain skepticism about diagnosis performed entirely by machines, and further tests are needed in the future. Because medical treatment is not only about diagnosis and treatment, but also involves the interaction between doctors and patients, especially doctors’ comfort to patients. The role of substitution." Chen Qiulin said.
In the future, "artificial intelligence doctors" may work together with human doctors. The reporter experienced this service model. On the Beijing Image Cloud Platform, the artificial intelligence system intelligently detects and recognizes nearly 9,000 CT images of lung nodules uploaded by 30 patients in primary hospitals, and marks the suspected nodules detected in the first round of screening as auxiliary diagnosis results , Provided to 4 radiologists for review. After reviewing, the doctor thinks it can be adopted, and then signs the report.
In this new service model, doctors still make the final decision. Some clinicians said: First of all, we must ensure that the artificial intelligence products are technically strong and give reasonable diagnosis suggestions; secondly, we must conduct training to change their concepts and adapt to the new service model. The doctor's approval and guidance will increase the patient's trust in the artificial intelligence system.
At present, the development of the medical artificial intelligence industry is still facing problems. "There is no interconnection and interoperability of data in various fields of medical care. The final result is big data, not big data. Medical artificial intelligence requires both medical talents and artificial intelligence talents. At present, relatively well-developed companies or very good projects are all It is promoted by talents in these two areas." Chen Qiulin said, because it involves the protection of personal privacy, it is necessary to define the property rights of medical data. Clear property rights are conducive to the realization of interconnection.
It is foreseeable that in the future, human beings will be inseparable from "artificial intelligence doctors." At that time, medical treatment was no longer "queuing medical treatment", but "second medical treatment", "precision medical treatment" and "personal medical treatment".
<INPUT type=checkbox value=0 name=titlecheckbox sourceid="SourcePh" style="display:none">