OFweekrobotinternet News This situation is commonplace: many top doctors are caught in daily trivial management and a large number of outpatient work all day long, and cannot spend their main time on patients who really need him; in addition, due to the relatively low level of information technology, doctors cannot effectively and timely obtain. The information of the patient cannot be intervened early, and the information of the patient and the prognosis cannot be effectively tracked.
In recent years, an opportunity for change is taking place. Since 2006, more and more companies have joined the “medical + artificial intelligence” field, with the largest proportion in 2016. In the past four years, this industry has developed rapidly.
This is mainly due to three factors. First, artificial intelligence technology has reached unprecedented new heights. Second, the combination of computer science and neuro-brain science has enabled humans to more and more understand the nature of intelligence. Thirdly, with the rapid development of the Internet, there are a lot of easily obtained massive data, which has greatly promoted the development of the industry.
Artificial intelligence can be applied in many medical fields. For example, in the imaging department, the annual growth rate of medical imaging data in China is 30%, while the growth rate of imaging doctors is only 4%. Therefore, artificial intelligence technology will free imaging doctors from heavy labor, and help them reduce the rate of misdiagnosis and improve the accuracy rate.
AI can also evaluate 8.2 million candidate compounds for drug development by screening treatments in databases of molecular structures. In 2015, Atomwise, based on the application of existing drug candidates, successfully found two drug candidates to control the Ebola virus in one day. A new drug, from drug discovery to FDA approval, takes an average of 97 months; and when entering the clinical trial stage, only an average of less than 12% of drugs are actually marketed. With the integration of artificial intelligence and machine learning, people are expected to reduce the cost of new drug research and development by an average of 28 billion US dollars per year.
However, there are still many difficulties in the application of artificial intelligence in the medical field.
At the data level, due to the fragmentation of interests and the lack of standardized standards, it is very difficult to interconnect data. In addition, due to information silos, many institutions have incomplete and severely unbalanced data.
How do we handle this situation? Transfer learning is suitable for dealing with the same data