Tech

When it comes to healthcare, AI has a long way to go


That’s because health data like medical images, vital signs, and data from wearables can change for reasons unrelated to a specific health condition, such as lifestyle. or ambient noise. The machine learning algorithms popularized by the tech industry are very good at finding patterns that they can discover shortcuts to “correct” answers that won’t work in the real world. Smaller data sets make it easier for algorithms to cheat that way and create blind spots that cause poor outcomes in the clinic. “Stupid community [itself] Berisha says we are developing models that work much better than the real thing. “It enhances the AI ​​hype.”

That problem has led to a striking and disturbing pattern in several areas of healthcare AI research, Berisha says. In studies that used algorithms to detect signs of Alzheimer’s disease or cognitive decline in voice recordings, Berisha and his colleagues found that larger studies reported more accuracy. less accurate than smaller studies – the opposite of what big data is supposed to provide. ONE review of the studies that tried to identify brain disorders from medical scans and again for studies attempting to detect autism with machine learning have reported a similar pattern.

The danger of algorithms that work well in preliminary studies but behave differently on real patient data is not hypothetical. A 2019 study found a system used on millions of patients to prioritize access to complementary care for those with complex health problems. put white patients before black patients.

Avoiding such biased systems requires large, balanced data sets and careful testing, but biased data sets are the norm in health AI research, due to health inequalities. past and ongoing health. ONE Stanford researchers’ 2020 study found that 71 percent of the data used in applied studies study carefully US health data comes from California, Massachusetts or New York, with little or no representation from the other 47 states. Low-income countries are barely represented in AI healthcare studies. Evaluate published last year of more than 150 studies using machine learning to predict diagnoses or disease processes concluded that most “show poor methodological quality and have a high risk of bias”.

Two researchers concerned about these shortcomings recently founded a nonprofit called Open Science Nightingale to try to improve the quality and scale of the data sets available to researchers. It works with health systems to manage collections of medical images and related data from patient records, anonymize them, and make them available to nonprofit research.

Ziad Obermeyer, co-founder of Nightingale and associate professor at the University of California, Berkeley, hopes providing access to that data will encourage competition leading to better outcomes, similar to that of collection. open image, how big. has helped drive progress in machine learning. “The core of the problem is that a researcher can do and say whatever they want in the health data because no one can check their results,” he said. “Data [is] locked up.”

Nightingale joins other projects aimed at improving healthcare AI by enhancing data quality and access. The Lacuna Foundation support the creation of machine learning datasets representing low- and middle-income countries and doing research in healthcare; One new project at University Hospitals Birmingham in the UK with support from the National Health Service and MIT are developing standards for assessing whether AI systems are anchored to unbiased data.

Mateen, editor of the UK report on pandemic algorithms, is a fan of such AI-specific projects but says AI’s prospects in healthcare also depend on the system. their modernized medical system. IT infrastructure is often shabby. “You have to invest in the root of the problem to see the benefits,” says Mateen.


Stories with WIRED are more amazing

.



Source link

news7g

News7g: Update the world's latest breaking news online of the day, breaking news, politics, society today, international mainstream news .Updated news 24/7: Entertainment, Sports...at the World everyday world. Hot news, images, video clips that are updated quickly and reliably

Related Articles

Back to top button