A new generation of medical AI product technology has entered a period of rapid development, and many companies have developed products that assist doctors in different departments. Since 2017, various medical AI companies have begun to recruit a large number of market personnel to speed up the layout.
In March 2018, the arterial network reporter learned from the director of the Department of Radiology of Shao Yifu Hospital, Hu Hongjie that the medical AI companies in their department had reached 10 companies. In April, the arterial network reporter learned from the director of the Department of Radiology of the First Affiliated Hospital of Chongqing Medical University, Lu Fajin, that the number of medical AI companies in the department has reached seven.
Many medical AI companies have gathered in hospitals. This is actually not an individual phenomenon. This is the case in most well-known top three hospitals. There are many products, but most of the doctors only use one product. In the face of numerous choices, what do the departments and doctors make to make their choices? Which medical AI products are actually used in hospital departments, used by doctors, and which are indiscriminate, just put a device or system hospital, and send a press release to promote financing.
The arterial network interviewed with these questions or learned about the directors of the Department of Radiology, the First Affiliated Department of Radiology, the Department of Respiratory Medicine of Beijing 301 Hospital, the Department of Blood Transfusion of Xiamen Second Hospital, and the Department of Ophthalmology of Shanghai North Hospital. The way AI is and the original intention.
Department of Blood Transfusion, Xiamen Second Hospital: Engineering personnel should study with doctors
Lai Dong, Director of Blood Transfusion Department and Central Laboratory, Second Affiliated Hospital of Xiamen Medical College
Lai Dong, director of the Department of Blood Transfusion and Central Laboratory of the Second Affiliated Hospital of Xiamen Medical College, told the Arterial Network reporter that they have two main starting points for medical AI products.
First, landing . All AI products, no matter how the company describes it, use machine learning, deep learning or other algorithms. Director Lai is most concerned about whether the product can solve the problems in the actual work, or can use the computer to focus on the clinical problems of doctors. Or the means of AI can be displayed to facilitate doctors to do analysis and research.
Second, the engineering staff should work with the doctor's professional maturity . Director Lai said that most of the data studied by the Department of Transfusion was non-image data.
When conducting research, they must first understand the variable data related to the disease. In addition to the highly specific data that doctors are aware of. Many variables, after the verification of big data and the diagnosis of the disease are related, the first thing doctors have to solve is to find the relevant variable data as much as possible.
Once the associated data is found, rules are created to structure the ordinary data so that it can be recognized and analyzed by the computer. This process involves the doctor's expertise, data structuring, integrity, and missing data values, all of which affect data quality.
After the data is processed, engineers can use AI technology to process and find the correlation between disease and data.
Compared with image data, the structuring and standardization of non-image data is more difficult, because today's image data is basically digitized after it comes out of the device. After the doctors mark it, it can be used as AI training. However, prescriptions and illnesses describe these non-image information. To use AI training, you must first do data structuring. This process is a test of the level of doctors and engineers.
Director Lai believes that engineers and doctors are a research and development team. When organizing the data, the two parties will carefully communicate and understand each other's needs and concerns. During the polishing stage, many small details and small problems will be solved.
These problems require the maturity of the medical AI companies and hospitals to be very high. At the same time, the doctor's energy is limited. A team of doctors generally chooses a company to carry out in-depth cooperation. The company will send engineers to the hospital to facilitate communication.
Director Lai said that their maturity with Yasen Technology is very high, which is why they can work with them for a long time.
Department of Radiology attached to a hospital: I want to enter the hospital to pass the test
Lu Fajin, Director of Radiology, the First Affiliated Hospital of Chongqing Medical University
The arterial network was learned from the director of the Department of Radiology of the First Affiliated Hospital, Lu Fajin. Their departments are currently in contact with seven companies engaged in medical AI research, but only three companies have entered their clinical polishing stage.
Director Lu Fajin told the arterial network reporter that due to the difference in medical AI product training data and hospitals, there is definitely a difference in performance between these products in the hospital. Therefore, it is necessary to enter the pre-clinical assessment.
The method of assessment is that the department establishes a test area according to the daily clinical work flow, and then verifies it with the clinical data of the attached hospital. Only the products that have passed the test can enter the clinical trial. Products that fail to pass can't go to the clinic, and feedback to the company allows them to correct themselves.
Through a period of use, first-line radiologists said that while paying attention to product accuracy, they also care about the convenience of product operation, and whether the product is consistent with the existing doctor's workflow.
The doctor's usual work is carried out in the hospital's information system. The radiologist has his own PACS system. In daily work, the PACS system is used to transmit, store and access image data. Therefore, the radiologist hopes that the medical AI auxiliary diagnosis system can also be embedded in the PACS system.
Those offline systems that require doctors to copy and transfer data will change the reading time from 10 minutes to 30 minutes, and such products will definitely be eliminated.
Department of Ophthalmology, Shibei Hospital, Jing'an District, Shanghai: Artificial application of artificial intelligence and innovation are equally important
Chen Jili, Director of Ophthalmology, Shibei Hospital, Jing'an District, Shanghai
Chen Jili, director of the Department of Ophthalmology, Shibei Hospital, Jing'an District, Shanghai, told the Arterial Network that Shibei Hospital is the northern regional medical center in Jing'an District. They also have their own considerations when choosing to use AI products.
First, AI products have become the important assistants of doctors in the future. It is a consensus and has become more and more widely used in hospitals . As a key medical subject in Shanghai, Shibei Hospital Eye Care, based on departmental development and discipline construction, needs to keep pace with the times, keep abreast of the latest medical technology trends, and apply useful technologies to the clinic and serve patients.
Second, Shibei Hospital is a comprehensive hospital. The doctors of the endocrinology department often ask the ophthalmologist to check the fundus of the diabetic patients to see if there is diabetic fundus lesions.
The biggest hazard of diabetes is various acute and chronic complications, especially sugar net disease, which leads to extremely high disability and blindness. However, if the fundus examination is performed regularly at the onset of the disease, the risk of blindness can be reduced by 94.4%. Therefore, early screening, early diagnosis, and early treatment are the key to retaining vision in patients with sugar net disease.
However, the current situation of distress is that the first diagnosis of diabetes patients is endocrinology, and many endocrinologists do not read the film at the moment, and cannot diagnose and judge the referral. Even if the endocrinologist can see the problem from the fundus image, the ophthalmology diagnosis report cannot be issued according to the law. Therefore, endocrinologists and ophthalmologists often consult together.
According to the normal process, doctors in the department should actively cooperate with the needs of colleagues, but the ophthalmologist itself is very busy. The endocrinology department lacks the ability to diagnose low images of the eye. Many of the eyes that require consultations are normal.
Therefore, Director Chen hopes that a reliable fundus screening product will help them carry out preliminary screening. After the system prompts abnormalities, the consultation will be carried out, so that ophthalmology can save a lot of manpower.
Third, there is a “three-year public health action plan†in Shanghai. One of the projects is sugar net screening. The project requires all parties to screen for sugar net disease and eye disease in the community, take a fundus photography in the community, and then upload. Data, let doctors in higher-level hospitals read films, health without lesions, no need to refer to the upper hospital ophthalmology, regular screening can occur, sugar network lesions need to be referred to higher-level hospital ophthalmology.
This sugar net screening work is very significant, but Director Chen said that the workload is also very heavy. In 2017, Shanghai has screened 180,000 diabetic patients. The upper hospital itself has very busy ophthalmologists, and the ophthalmologists have increased their readings. The workload caused the readings to fail to produce results in a timely manner. The patient waited for a long time, and the difference in clinical experience of the reading doctors also led to uneven reading materials.
In fact, the effective way to solve the above problems is to let the artificial intelligence auxiliary diagnosis system of the fundus solve the preliminary screening work. At present, there are many hospitals and research companies that develop artificial eye-assisted diagnostic systems for fundus cameras, but none of them have passed the national CFDA certification, so they can not be used in clinical practice.
At present, there are many technology companies in China that have launched the fundus AI-assisted diagnostic software. Director Chen has chosen a partner to have his own standards.
Director Chen said that on the one hand, the development company's R&D team should be strong enough to continuously optimize and nurture its own AI system, and the product is good enough, and the correctness and sensitivity of fundus lesion recognition is high enough. The diagnostic system is what is needed in the future.
On the other hand, they need more than just clinical applications. They also hope to cooperate with these technology companies to develop research and develop more AI products. Currently, they are developing AI-assisted diagnostic systems for common fundus cameras. Director Chen hopes to make a contribution to the AI-assisted diagnosis system for ultra-wide-angle laser scanning cameras .
The general fundus camera has a shooting angle of 45 degrees, and the super wide-angle fundus camera has a shooting angle of 200 degrees. The super wide-angle fundus camera has a wide shooting angle, and it can find more hidden problems in the fundus. The future will be a development direction.
In this cooperation, City North Hospital and Airdoc hit it off. Airdoc has applied deep learning technology to the medical field since 2015. It is the first automatic identification algorithm for artificial intelligence diabetes eye disease in China, and the first company to promote artificial intelligence technology to the market. Therefore, City North Hospital and Airdoc hit it off.
With the support of Airdoc, in October 2017, the application of the ultra-wide-angle fundus laser imaging system combined with artificial intelligence image analysis technology in the development of fundus disease-assisted diagnosis was jointly obtained by Shanghai North Hospital and Shanghai Xinhua Hospital. The action plan is supported by 500,000 yuan. The research is currently on schedule, and in the near future, the AI-assisted diagnostic system for super wide-angle fundus cameras developed by themselves will be available.
In February 2018, with the full support of Airdoc, the artificial intelligence screening construction project of the northern medical community in Shanghai Jing'an District was officially launched. This is the first time that medical artificial intelligence technology has been applied in the clinical field of ophthalmology in Shanghai, north of Jing'an District. In the future, diabetic residents in the area can conduct preliminary screening of artificial intelligence of diabetic retinopathy at the community health service center. The fundus photos will be transmitted to the Ophthalmology Department of Shibei Hospital through the Ophthalmology Big Data Platform of the National Science and Technology Information Center in time. People with diabetes in the north do not need to go to the secondary and tertiary hospitals for ophthalmology to make routine regular diabetic fundus examinations.
Director Chen explained that the sugar net artificial intelligence screening system and the Shanghai three-year public health action plan sugar net screening are not contradictory, they are a good supplement, this work screening target community health service center visit In patients with diabetes, when the patient dispenses the medicine, the general practitioner can suggest that they take a photo of the fundus. The Airdoc fundus-assisted diagnosis system can immediately give a result of the fundus lesion. If there is a fundus lesion, it can be referred to the green channel. City North Hospital ophthalmology review to give the correct treatment.
Director Chen said that the ultimate goal is to establish a comprehensive screening and prevention service system for diabetic retinopathy called “screening-discovery-referral-follow-health management†based on artificial intelligence technology.
This sugar net and chronic disease screening system based on artificial intelligence and big data platform covers concepts such as disease discovery, referral, treatment, follow-up, and health management, and can effectively solve the problems of sugar net and chronic disease screening. Director Chen said that in the future, this service system will be promoted and used throughout the city and the country.
Department of Radiology, Run Run Shaw Hospital: 10 companies have not fully covered the hospital's AI R&D
As a hot spot for medical AI innovation applications, many companies have branches in Hangzhou, or Hangzhou hospitals as the first target.
Shao Yifu Hospital is a public top three hospital that has passed the JCI International Hospital for four times and is highly concerned by medical AI companies. The arterial network reporter learned in the field interview that since 2016, 10 medical AI companies have cooperated with the radiology department of Run Run Shaw Hospital to polish medical AI products.
In these cooperative AI companies, the depth of cooperation varies. Some companies only put the products in the hospital for doctors to use, the effect is difficult to assess; some are installed in the hospital's PACS system with auxiliary diagnostic software, the doctor is relatively convenient to use.
But what's interesting is that most AI companies cut through the lung nodule project and have relatively few R&D in other areas. The R&D needs and clinical needs that radiology and other departments could have solved with AI have not been met. They can only develop or cooperate with universities, and even some doctors write programs on their own.
For example, automatic grading around the liver, pre-predicting early recurrence of liver cancer, quantitative analysis of emphysema subtypes and correlation with lung function, and simple pre-diagnosis based on imaging omics/machine learning methods system. (For details, see "Expectation and Confusion, Shao Yifu Hospital's Innovation and Application in the Field of Medical AI")
The research projects that doctors organize themselves must be the problems they see in the clinic. These are real clinical needs that need to be addressed, not just the lung screening products. Even lung screening products have not been perfected.
Zhang Xiaoyu, deputy director of the Department of Radiology at the Run Run Shaw Hospital, said that the current stage of AI is AI and PACS is PACS. The two have not yet fully realized seamless coordination. This makes the doctor seem to be not humanized during the operation, which will affect the enthusiasm of the doctor to some extent. Seamlessly connecting the AI ​​system to the PACS system will also be the next step in development.
Now, before submitting the report, many doctors will choose to refer to the report issued by the AI ​​system, compare it with their own and check for missing. The question of which product to use is based entirely on the accuracy and ease of operation of the product.
In addition, there are many diseases that radiologists need to screen, such as thyroid nodules, breast nodules, liver mass, and prostate and other abnormal lesions. The reporter believes that many developers in the field of diseases are often few. If there are collaborators who are reliable and practical, doctors will consider cooperating with them.
Department director and doctor four choice basis
By communicating with the director and doctor, we summed up the basis for the four-point department director and doctor to choose medical AI.
First, the team must be reliable, not just telling stories. It is the scorpion that is marathon, showing the product model to the hospital, and understanding the products and diseases.
Second, the accuracy of the product is the premise of entering the clinical trial. No matter which hospital or doctor, they have high requirements on product performance, but the products can only be used in hospitals.
Third, the products must be innovative. Nowadays, there are many medical AI products, but the homogenization is serious. In fact, there are many scientific research needs and clinical needs of doctors. As long as you grasp a point, you may let the products fall in the hospital.
Fourth, engineers must work with doctors to polish products.
Fifth, AI products are embedded in the doctor's workflow, so don't bring extra trouble to the doctor. In the imaging department, the AI ​​system can be embedded in a PACS or a doctor's imaging workstation.
Rhodium powder Rh≥99.95% Properties: silver-white metal, extremely hard, wear-resistant, also has considerable ductility Application: It can be used as raw material for electrical instrumentation, chemical industry and manufacture of precision alloys. Rhodium powder is based on the widespread use of ruthenium in the industrial industry. Since rhodium is a rare metal required by the industry, the industry price is slightly higher than that of general non-ferrous metals. Rhodium, one of the rare elements, has various uses. Rhodium can be used to make hydrogenation catalysts, thermocouples, platinum-rhodium alloys, etc. It is also often plated on searchlights and reflectors, and is also used as a polishing agent for gemstones. and electrical contact parts.
precious metal
XI AN RHINE BIOLOGICAL TECHNOLOGY CO.,LTD , https://www.rhinebioteches.com