Release date: 2017-02-10
Artificial intelligence is a good helper for doctors, and everyone is familiar with this. There have been many medical robots for screening and assisted decision making, such as IBM's Watson and Deepmind's AIphaGo, both of which are based on big data. However, for rare diseases, it is difficult to obtain high-quality data, which raises a question: Can artificial intelligence be applied in screening and assisted diagnosis of rare diseases?
Inspired by the paper published by Google DeepMind in 2015, Lin Haotian, a post-80 ophthalmologist at the Zhongshan Ophthalmology Center of Sun Yat-Sen University, China, and his colleagues came up with the idea of ​​creating an artificial intelligence platform to mine their clinical data on congenital cataracts and then to screen them. And the purpose of aiding diagnosis. Together with Professor Xi Xiyang from Xi'an University of Electronic Science and Technology, I used the champion model of ILSVRC 2014 (ImageNet Large Scale Visual Recognition Challenge of 2014) to establish a deep learning model for identifying congenital cataracts. This model is considered to be dominant in the field of image recognition and can be used. Training and classification), named CC-Cruiser. The results of this research were published in the journal Nature Biomedical Engineering on January 30, 2017 [2].
Lin Haotian
The congenital cataract was chosen as the subject of this study because it causes blindness and visual impairment, which occurs before or after birth, or occurs during childhood. The incidence rate in China is 0.05%. In addition, it is a typical rare disease that combines the characteristics of chronic and acute diseases and is suitable for test cases.
Trained CC-Cruiser's photo collection, the researchers used a partial picture of a routine check from the Chinese Ministry of Health's Children's Cataract Program (CCPMOH), including 410 images of children with congenital cataracts of varying severity, 476 normal children Eye image. All images were independently classified and described by two experienced ophthalmologists, and the third ophthalmologist advised on disagreements. The three human physicians have not been exposed to CC-Cruiser.
Â
After completing the training, CC-Cruiser basically has three functions:
- Screening for patients with congenital cataracts
- Risk assessment of patients, the assessment has three main indicators: opaque area, depth and position of the lens
- Assist ophthalmologists in making treatment decisions
Â
The researchers conducted five tests on CC-Cruiser. In the computer simulation test, the accuracy rate of CC-Cruiser screening for congenital cataract patients was 98.87%. Among the risk assessment functions, the accuracy of the three indicators (opaque area, depth and position) was 93.98%, 95.06% and 95.12%. In assisting decision making, the accuracy rate for providing advice to ophthalmologists is 97.56%.
To further explore the versatility and practicality of CC-Cruiser, the researchers selected three non-ophthalmic hospitals for testing, two hospitals in Guangzhou and one in Qingyuan. Because the researchers hope that CC-Cruiser will ultimately help these hospitals that lack on-site ophthalmologists. In 57 children's eye pictures, the accuracy of screening was 98.25%. The accuracy of the three indicators (area, depth, location) of the risk assessment were 100%, 92.86% and 100%, respectively. The accuracy of the decision-making decision is 92.86%.
The picture shows the comparison test between CC-Cruiser and human physician.
CC-Cruiser also performed very well in comparison tests with human ophthalmologists. The panel consists of three ophthalmologists, an ophthalmologist with ten years of experience, a competent doctor who has completed ophthalmic clinical training and specific training, and a novice doctor who has completed theoretical studies and started clinical practice. In 50 images, CC-Cruiser identified all congenital cataract patients. The three ophthalmologists made mistakes in the third picture - mistakenly diagnosed the high-light area of ​​the picture as a congenital cataract. CC-Cruiser performed well in risk assessment and decision-making, and gave correct treatment recommendations to all patients who needed surgery. Therefore, the researchers believe that CC-Cruiser can be called a "qualified ophthalmologist."
However, being a "qualified doctor" is not enough. At present, the management of rare diseases in China is relatively backward, mainly because of the high cost and scattered geographical area, and it is difficult for patients to enjoy professional high-quality care. China has a large population, and rare disease management often has problems such as missed diagnosis or misdiagnosis and improper treatment decision-making. This is far from the goal of “precise treatment†that has been advocated. To this end, the researchers moved CC-Cruiser to the "cloud" for the management of medical services for congenital cataracts.
The researchers established the CC-Cruiser website (https://). Unlike the Watson robots for hospitals, regular users can access the site through registration. After entering the website, users can upload their own eye pictures, and the website will immediately give a diagnosis. The diagnosis results include: whether there is disease, size, depth, location, treatment advice (surgery or follow-up). At the same time, the website provides 50 typical cases for users who want to test the CC-Cruiser program. These cases cover most of the clinical situation, normal conditions, cataracts without surgery, transparency of different severity, should Accept high-risk cases for emergency intervention.
Interface of CC-Cruiser's website (results after sample test with singularity cake)
After the user's consent, the user's clinical information (currently image) and demographic information will be uploaded to the website and saved to the database. If CC-Cruiser recommends that the patient undergo surgery, the CCPMOH doctor will be notified immediately for further confirmation. To prevent misdiagnosis and missed diagnosis, CCPMOH's doctors will check all uploaded cases, including those that CC-Cruiser judges to be normal and sick. The patient can then communicate with the doctor and the website provides an email and daytime phone service to confirm if surgery is needed.
With a curiosity, after entering the CC-Cruiser website, I first used a mobile phone to take a picture of my own eyes. The result is that I am not sick, very relieved! Then use the pictures in the sample test on the website, the result is sick. I also found a picture of the eye of a congenital cataract on the Internet. The result was sick. Then, because the website did not require the image format, the singularity of the article took the test of the cover picture of the article and the result was still ill. Finally, the singular cake was casually looking for a picture of a lip that was not right, it was the picture below, and the result was still ill... It seems that the researcher needs to have a picture format uploaded on the website.
In an interview with IEEE Spectrum, Lin Haotian said that doctors should make full use of artificial intelligence advice to prevent potential mistakes and supplement their own judgments, but at the same time, the interaction between doctors and patients is essential for treatment. For machines, it is very challenging to imitate people's emotions. The interaction between the doctor and the patient will be the last bastion of human wisdom.
The research team hopes that in further clinical trials, doctors in non-ophthalmic hospitals can use the program to screen patients and send them to a specialized medical center. The patient can eventually use it. Just as AI can be understood not only as Artificial Intelligence, but also as Augmented Intelligence. Lin Haotian believes that the ultimate goal of artificial intelligence is to use it to combine with human capabilities to create a better world.
references:
[1]http://spectrum.ieee.org/the-human-os/biomedical/diagnostics/ophthalmologists-vs-ai-its-a-tie
[2]Erping Long, Haotian Lin, Zhenzhen Liu, Xiaohang Wu, Liming Wang, Jiewei Jiang, Yingying An, Zhuoling Lin, Xiaoyan Li, Jingjing Chen, Jing Li, Qian Zhong Cao, Dongni Wang, Xiyang Liu, Weirong Chen & Yizhi Liu. An artificial intelligence platform for the multihospitalcollaborative management of congenital cataracts. Nature Biomedical Engineering [J] doi:10.1038/s41551-016-0024
Source: Singularity Network
Squid Flower,Frozen Calamar Flower,Frozen Octopus Flower,Frozen Pineapple Cut Squid Flower
Zhoushan Haiwang Seafood Co., Ltd. , https://www.haiwangseafoods.com