How far is the artificial intelligence + smart medical robot doctor from us?

Based on deep learning, Guangzhou Women and Children's Medical Center has developed an artificial intelligence system that can diagnose diseases such as eye diseases and pneumonia. The research results were published in the world's top journal “Cells” on February 23.

This artificial intelligence result can give doctors a diagnosis suggestion based on the image data and explain the basis for the judgment. The comparison experiment found that the accuracy rate of the system in the diagnosis of eye diseases was 96.6%; the accuracy rate was 92.8% in distinguishing between pneumonia and health status, which is comparable to the expert doctor with more than ten years of experience.

How big is the skill|Precision medication, second-order judgment

Pneumonia is the leading cause of death among children worldwide due to infection. Finding a pulmonary nodule from a chest CT, a trained doctor takes an average of 3-5 minutes, while relying on artificial intelligence takes only 3-5 seconds.

This is the artificial intelligence platform developed by the Guangzhou Women and Children Medical Center and the University of California San Diego team led by Professor Zhang Kang.

Not only fast, but more important is accurate. The key factor in determining the prognosis of pneumonia is whether it can be accurately administered according to the type of pneumonia. Traditional methods based on blood culture, sputum culture, biochemical detection, etc., are difficult to judge quickly and accurately. The artificial intelligence platform can achieve accurate second-order determination of the type of pneumonia in children based on the chest X-ray of children.

This enables the use of artificial intelligence to accurately guide the rational use of antibiotics, and the platform can be free from the limits of hospitals and regions, achieving wide coverage of community health care, family doctors, and specialist hospitals, providing accurate diagnosis of pneumonia, an area where antibiotics are abused. Medication program to avoid the abuse of antibiotics and promote the rehabilitation of severe pneumonia in children.

The artificial intelligence platform has important clinical significance. People expect artificial intelligence with higher efficiency and precision to become a good helper for doctors. Artificial intelligence will make a difference in the screening and prevention of pre-diagnosis diseases, medical image-assisted diagnosis, analysis of test results, surgical assistance, medical follow-up after treatment, chronic disease monitoring, rehabilitation assistance, and health management. It will even bring changes to basic research aids, drug development, gene screening analysis, and medical training.

"Now our artificial intelligence platform can prevent more people from being identified, early diagnosis and early treatment anywhere in the world without being restricted by the region." Joining the Guangzhou Women and Children Medical Center Genetic Testing Center in 2016 Zhang Kang, a professor at the Shiley Eye Institute at the University of San Diego, said.

Trustworthy? High accuracy, process visible

Some people say that artificial intelligence is reliable? Is it safe to give your life to the robot?

The research team cut through the two diseases of macular degeneration and diabetic retinal macular edema, allowing this artificial intelligence system to continuously learn eye optical coherence tomography images. After learning the image data of more than 200,000 cases, the accuracy of the platform for the diagnosis of macular degeneration and macular edema reached 96.6%, and the sensitivity reached 97.8%. Compared with the results of the five ophthalmologists, the confirmation platform can reach the level of a trained ophthalmologist and determine whether the patient should receive treatment within 30 seconds.

The reporter learned that this artificial intelligence system has deep learning ability. AlphaGo, autopilot, and other applications that are well known are based on deep learning techniques.

In this research and development process, the research team applied a new algorithm based on the migration learning model, which not only greatly improved the learning efficiency of artificial intelligence, but also helped achieve the goal of “one system to solve multiple diseases”.

“Traditional deep learning models typically require millions of high-quality, same-type annotation data to achieve more stable and accurate output, but in reality, collecting millions of high-quality annotation images for each disease is almost no. It is possible to realize that the wide coverage of artificial intelligence in the field of medical imaging is difficult to achieve." Zhang Kang introduced. Therefore, the existing medical artificial intelligence generally only one system can target one disease.

Relatively speaking, this artificial intelligence platform based on the migration learning model requires very little data, and researchers can complete a cross-path migration with only a few thousand.

For example, in this study, based on the artificial intelligence system trained by 200,000 eye image data, the research team used only 5,000 chest X-ray images to construct an artificial intelligence image diagnosis system for pneumonia. A differential analysis and second-order determination of the pathogen type of childhood pneumonia was achieved. After testing, it has an accuracy of 92.8% and a sensitivity of 93.2% when distinguishing between pneumonia and healthy state. In distinguishing between bacterial pneumonia and viral pneumonia, the accuracy is 90.7% and the sensitivity is 88.6%.

In addition, in the past, research and products that rely solely on deep learning techniques only gave results in the reports, but did not list the reasons and processes for judgment. This "black box" type of diagnosis, even if the accuracy is high, the doctor does not Dare to use. This artificial intelligence platform has overcome this limitation to a certain extent, and people "know it, but also know why."

The research team used the idea of ​​occlusion testing. Through repeated learning, practice and improvement, the platform can display which region of the image to obtain the diagnosis result, and to some extent give the reason for judgment, thus making it more credible. .

Prospects have geometry | system assessment, assisted decision making

Artificial intelligence diagnoses diseases so efficiently, how far is the robot doctor from our lives?

Zhang Kang said that their artificial intelligence systems are currently undergoing small-scale clinical trials in eye clinics in the United States and Latin America. In addition, in subsequent studies, they will further increase the number of data learning templates, increase the number of diagnosable diseases, and further optimize the system.

As early as 2015, Guangzhou Women and Children Medical Center launched the “Mim Bear” smart family research and development project based on medical big data and integrated artificial intelligence frontier technology.

"This family member has four bears, a fever bear, an image bear, a guide bear, and a nutrition bear." Liang Huiying, director of the clinical data center of the hospital, said that "Fever Bear" is based on the common fever-related diseases in children. Authoritative guides, expert consensus, more than 2 million copies of knowledge documents such as massive medical records, combined with multi-source heterogeneous data integration technology, natural language processing technology and machine learning algorithms, after a year of training, have been successfully targeted at 24 kinds of children An accurate auxiliary diagnosis for fever-related diseases is an intimate assistant to the outpatient doctor by seamlessly embedding the electronic medical record system.

The image bear is based on "chest X-ray film + microbial culture detection big data", using deep learning algorithm to intelligently identify the microbial infection status of pneumonia (bacterial, viral, mixed infectivity), providing decision support for the precise application of antibiotics It has been applied to the doctor's auxiliary diagnosis. The data and technology formed in its practice have become an important foundation and component of the scientific research results of artificial intelligence systems.

The other two "bears" are also thriving and will be expected to meet the public in the near future.

The results of medical artificial intelligence research published in the journal Cell are regarded as a new starting point by the Guangzhou Women and Children Medical Center. Center Director and Dean Xia Huimin said, "The ultimate goal of the artificial intelligence platform is to integrate text-based medical record data, fully structured laboratory inspection data, image data, photoelectric signals and other multi-media data to simulate clinicians' systematic assessment of patients' conditions. To provide a comprehensive and complementary decision-making for medical staff, not just to provide a single-sided decision-making decision for a video doctor or a medical technician."

“Therefore, the platform is still intensifying.” Xia Huimin said, for example, in the field of intelligent discriminating types of childhood pneumonia, the team is systematically reading X-ray films, increasing laboratory tests and clinical symptoms. More accurately determine the type of pathogens in children with pneumonia.

"I hope that in the near future, this technology can be applied to primary health care, community health care, family doctors, specialist hospitals, etc., to form a wide-ranging automated triage system." Xia Huimin said.

This set of artificial intelligence is so "smart"

This artificial intelligence adopts the migration learning algorithm, which is to transfer the trained model parameters to the new model to help the new model training, that is, to use existing knowledge to learn new knowledge and find existing knowledge and new knowledge. The similarity between the two is dictated by the idiom.

For example, if you have learned to play Go, you can learn chess in analogy. If you play basketball, you can learn volleyball analogously. If you have Chinese, you can learn English, Japanese, and so on. How to reasonably find commonality between different models and then use this bridge to help learn new knowledge is the core of “migration learning”. Migration learning is considered an efficient technique, especially when faced with relatively limited training data.

Taking medical image learning as an example, the system will recognize the characteristics of the image in the pre-system, and the researchers will continue to import the network system containing the first layer of image similar parameters and structure, and finally construct the ultimate level.

Innosilicon Mining Machine

Innosilicon is a worldwide one-stop provider of high-speed mixed signal IPs and ASIC customization with leading market shares in Asian-Pacific market for 10 consecutive years. Its IP has enabled billions of SoC's to enter mass production, covering nodes from 180nm to 5nm across the world`s foundries including: GlobalFoundries, TSMC, Samsung, SMIC, UMC and others. Backed by its 14 years of technical expertise in developing cutting-edge IPs and ASIC products, Innosilicon has assisted our valued partners including AMD, Microchip and Microsoft to name but a few, in realizing their product goals.
Innosilicon team is fully devoted to providing the world's most advanced IP and ASIC technologies, and has achieved stellar results. In 2018, Innosilicon was the first in the world to reach mass production of the performance-leading GDDR6 interface in our cryptographic GPU product. In 2019, Innosilicon announced the availability of the HDMI v2.1 IP supporting 4K/8K displays as well as our 32Gbps SerDes PHY. In 2020, we launched the INNOLINK Chiplet which allows massive amounts of low-latency data to pass seamlessly between smaller chips as if they were all on the same bus. With a wide range of performance leading IP in multiple FinFET processes and 22nm planar processes all entering mass production, Innosilicon's remarkable innovation capabilities have been proven in fields such as: high-performance computing, high-bandwidth memory, encrypted computing, AI cloud computing, and low-power IoT.

Innosilicon Mining Machine:Innosilicon T2 Turbo 25T,Innosilicon T3+ 52T,Innosilicon T3 43T,Innosilicon T2 Terminator,Innosilicon T3 39T,Innosilicon T2 Turbo+ 32T

Innosilicon Mining Machine,T3 Pro Miner,T3 Pro Innosilicon,T3 Pro 67T Innosilicon

Shenzhen YLHM Technology Co., Ltd. , https://www.ylhm-tech.com