Zeng Mu Xin, director of national chiao tung university Institute of Data Science and Engineering, has pr
Zeng Mu Xin, director of national chiao tung university Institute of Data Science and Engineering, has promoted the development of artificial intelligence with big data, algorithms and computing power. At first, the way of AI development and operation was that experts defined knowledge and machines simulated experts' ideas to solve problems. Later, machine learning methods appeared, and a new generation of algorithms such as deep learning and high-speed computing GPU were applied to realize an AI system like Alpha go. In recent years, AI medical care has developed rapidly, and the key is big data. The integration of medical data will produce great power. ? Application examples of AI medical care at home and abroad?
AI medical care is regarded as one of the most promising areas by the venture capital circle. In the past two years, nearly 100 new companies have emerged in North America, covering many fields, such as medical imaging and health risk prediction systems. After Google's AlphaGo won the Go competition, it began to cooperate with many hospitals, such as applying it to the diagnosis of ophthalmoscope retinopathy. The huge information technology center of China Institute of Technology and Jiaotong University also cooperated to develop an automatic recognition system for retinopathy, collected tens of thousands of images and borrowed Deep. The integration of medical images and electronic medical record data can further improve the accuracy. A system like this can be popularized and applied in front-line family medical clinics to achieve the purpose of early diagnosis and early treatment. At the same time, the data of laboratory physiological tests can be connected in series, and the microvascular analysis of retinal fundus images can be used to estimate whether there is cardiovascular disease. ? In addition to technology research and development, Israel's Zebra company has created a new business model in the field of AI medical care, allowing users to get a second medical order service by uploading medical images. Zebra can not only collect a large number of anonymous image data on a global scale through this service model, but also users are willing to pay to upload their own data to obtain analysis services. ? Does AI Medical Find More Early Lesions and Improve the Chance of Early Diagnosis?
The ability of image recognition based on artificial intelligence has been greatly improved, and it has recently been applied to telemedicine and nursing service. For example, the Aicure platform in the United States uses facial recognition and motion sensing technology to automatically observe patients' medication and reaction behavior through smart phones, helping doctors decide prescriptions and realize remote treatment. ? The iDeepCare project of the Ministry of Science and Technology, led by Professor Zeng, combines deep machine learning and massive data analysis technology to develop an intelligent deep health medical system, and cooperates with domestic medical institutions and experts to collect all kinds of biomedical raw data and establish models, which can be applied to precision medical care, preventive health care, personalized medical care, risk prediction and so on. For example, the iDeepCare project and three-in-one colonoscopy images have established and developed an AI identification model for colorectal polyps, which can help doctors identify the nature of polyps with an accuracy rate of 96%. In addition, we also cooperated with Beirong arrhythmia early warning system to predict the ECG manifestations of Bruguida syndrome with high risk of sudden death. The accuracy of general internal medicine trainers is only 47.5%, and it can reach 75% through AI machines. ? Through technical integration and analysis of large databases such as hospitals and medical insurance data, we can find the early lesions of various diseases and dig out many useful or previously unknown markers, which is helpful for doctors to make early diagnosis. ? Crossing the threshold of scientific research, will AI share weal and woe with mankind?
Any scientific research will have challenges. The biggest challenge of AI medical care lies in the quality of data, how to avoid the threshold of cross-domain cooperation such as "garbage in, garbage out", how to design business models, and how to catch up with market demand. Professor Zeng Mu Xin said that AI medical research can be based on existing data and innovative ideas can be found through cross-disciplinary brainstorming. At present, the most striking is the application of real-time dynamic analysis. In the future, the cooperation between human beings and AI will greatly improve the medical accuracy. AI will not replace human beings, but become an important helper for human beings. ?