Seeding Machine,Seed Drill,Seed Planter,Drum Seeder Taizhou Yingtian Agricultural Machinery Manufacturing Co., Ltd. , https://www.sakuradaagc.com
According to foreign media reports, with the development of artificial intelligence technology, its application in the field of medical health has become more and more extensive. However, due to the application of artificial intelligence technology such as machine learning technology, the medical products based on it are also in constant development, which is in conflict with the FDA's current censorship system. How to supervise new technologies while ensuring product safety is a key issue for the FDA to address.
At the technology conference in Toronto this fall, Anna Goldenberg, a star expert in computer science and genetics, explained how artificial intelligence will change the field of medicine. At present, artificial intelligence algorithms based on machine learning can identify the symptoms of skin cancer based on patient photos, and the recognition rate is higher than that of doctors. In addition, in the detection of cardiac rhythm abnormalities by analyzing electrocardiogram, the recognition rate of artificial intelligence is also higher than that of cardiologists. In Jinberg's lab, artificial intelligence algorithms can even be used to identify subtypes of brain cancer that have been blurred so far, to estimate the survival rate of breast cancer patients, or to reduce unnecessary thyroid surgery.
We have already smelled some about the future of artificial intelligence. According to data from the McKinsey Global Institute, large technology companies invested $30 billion in artificial intelligence in 2016, and artificial intelligence startups invested $9 billion. Many people are familiar with machine learning. This technology will automatically improve the analytical model through new data input into the computer, analyze the relationship between new trends and things to optimize predictive power. Machine learning allows Facebook to analyze the connections between friends and relatives, and allows Google to recommend where users have lunch. These features are very useful, but they are pediatric compared to the new way machine learning changes health care in the next few years.
The development of science and technology is unstoppable, and the flow of funds is unstoppable. But for the application of artificial intelligence in the field of health care, the obstacle that cannot be avoided is the supervision of the US Food and Drug Administration FDA.
Each professional field has its own hero. In the US Food and Drug Administration's censorship, the greatest hero was Frances Oldham Kelsey, who in the 1960s refused to allow permission to stop Kevadon from entering the US market. The drug can significantly inhibit pregnancy adverse reactions in pregnant women. But with widespread use in other countries, doctors have discovered that this drug, called thalidomide, can cause neonatal malformations. Kelsey still defends the principle of safety in the face of strong pressure from pharmaceutical companies, which has also spurred the FDA to introduce a rigorous evaluation model for drugs, hospitals and medical software. This model has been in use today.
At the heart of this evaluation model is the assumption that any product should undergo rigorous clinical testing, production, and marketing processes, and be used in a fixed form. That's why many people use sphygmomanometers in pharmacies that look the same as they did ten years ago. The FDA's assessment model requires a complete approval process, and costs and time costs increase accordingly.
However, building and freezing models is not usually a viable way of artificial intelligence software development, especially in machine learning. These artificial intelligence systems are essentially a meta-algorithm, and whenever new data is added, there is a new output, which means that virtually every day, there are countless new "medical devices." (A non-medical field example is how to better recognize the voice recognition process of a user's voice through training.) This phenomenon creates a gap between artificial intelligence medical software development and traditional regulatory systems.
For example, Cloud DX: This Canadian company uses artificial intelligence to analyze the audio waveform of human cough, which can detect asthma, tuberculosis, pneumonia and other lung diseases. In April, the XPRIZE Foundation in California named Cloud DX the bold innovator "Bold Epic Innovator" in its organization's Qualcomm Tricorder competition, and R&D participants were asked to develop a patient to measure their vital signs. Universal device. The company received a $100,000 bonus, but the FDA has not yet approved the product for clinical use. Under the current review system, obtaining such approval is very difficult.
This helps explain why many medical software innovators are looking for other creative ways to bring their ideas to market. Robert Kaul, founder and CEO of Cloud DX, said: "Technologies like Google are far from meeting the FDA's diagnostic criteria for clinical trials. "This could be a nightmare, and innovations are simply not adapted to this review and slow. jobs. He pointed out that as the benchmark for FDA equipment standards, even if you get the most basic ISO 13485 certification, you will need to spend seven figures of money and two years. "How many investors will give you so much money, so that you Can you reach the starting line?"
Former Google executive Vic Gundotra said: "My company is currently developing a software that can detect heart disease at an early stage, and 20% of the company's employees are dedicated to dealing with regulatory issues. "At Google, sometimes we will quickly decide some Things can be shipped even after six weeks. So when I got here, we had a breakthrough. I would say, “Can we ship quickly?†They will say, “Two yearsâ€, the original “quick-breaking†digital creed is completely unfeasible now.
Kaul has hope, because he believes that XPRIZE's help helps him to quickly pass the Cloud DX system through FDA review. Like everyone else, he recognizes that the FDA has an important role to play in protecting patients from false claims and dangerous products. He even saw the expansion of the institution. He said: "For those very few companies that have achieved their review goals, they have a huge competitive advantage. We don't have to worry about the usual situation: two people at Stanford University have developed a new application in the garage, immediately Take away all our business. We only need to worry about big companies, they may buy our technology instead of competing with us."
A long-standing complaint about the FDA review process is a cliché. Scott Gottlieb, the current head of the FDA, criticized the agency's slowness in reviewing life-saving drugs for children with Hunter syndrome five years ago. With the development of technology, this contradiction has become more acute as software algorithms have become an important part of the health system.
Canadian start-up company WinterLight Labs is developing a machine learning software that can detect various forms of cognitive impairment by analyzing patient speech segments, including early Alzheimer's disease. The technology is currently being tested in collaboration with a number of care providers. But Liam Kaufman, the company's chief executive, is not sure if or when his technology will pass the FDA's review, in part because it's unclear whether the FDA will ask him to freeze his product under defined conditions. His alternative plan is to sell the product as a screening tool, just to let the user know when to consult the doctor.
Bakul Patel, director of the FDA's new Digital Health Associate Center, recently launched a pilot program called FDA Pre-Certification, which, according to the announcement, will eventually allow agency officials to focus their review on "software developers or digital health technology developers, not on On the product. The nine companies selected for this initial plan include Apple, Fitbit, Samsung and several smaller companies. The public statement seems to imply that it is possible that one day these pre-certification companies can optimize their software products without having to ask for FDA approval at each iteration?--Although Patel has a strong background in business and technology, at this point Still ambiguous.
“We are improving this area,†he said. “The traditional review model is our current method. But we still can't verify artificial intelligence such as machine learning. So this question is the same research as the regulation itself: How do you adjust the performance of today's machines to be as good as yesterday?â€
Now, Patel is "crazy recruiting" related professionals to improve the digital strength of the FDA. But finding the right person is not easy, because the field of artificial intelligence is too hot. As reported in the New York Times Weekly, current experienced AI professionals are serving large technology companies with annual salaries of more than $300,000, which is much higher than what the US Food and Drug Administration can pay.
“Yes, it’s hard to recruit suitable candidates in the field of artificial intelligence.†Patel acknowledges this. “We have some understanding of these technologies. But we need more professionals. It will be a challenge.â€
Original title: Artificial intelligence brings new problems to the supervision of the Food and Drug Administration
The development of science and technology can't stop artificial intelligence from bringing new challenges to food regulation
[ China Pharmaceutical Network Technology News ] The development of science and technology is unstoppable, and the flow of funds is unstoppable. But for the application of artificial intelligence in the field of health care, the obstacle that cannot be avoided is the supervision of the US Food and Drug Administration FDA. The traditional review model is the current method. But it is not yet possible to verify artificial intelligence such as machine learning. So this question is the same research as the regulation itself: How do you adjust the performance of today's machines to be as good as yesterday?
(Science and technology development can not stop artificial intelligence to bring new challenges to food regulation)