Wednesday, October 23, 2024

We can vet your bot, for a fee

The ideas and innovators shaping health care
Oct 23, 2024 View in browser
 
POLITICO Future Pulse Newsletter Header

By Ruth Reader, Erin Schumaker and Daniel Payne

TECH MAZE

The Nvidia logo is shown on a company building.

Nvidia is among the firms offering help vetting AI tools. | Jeff Chiu/AP

Everyone’s got a platform to vet your AI.

At least that’s how it seemed at this week’s HLTH health tech conference in Las Vegas, Ruth found.

Absent government regulation of many AI tools in health care, the private sector’s filling the void.

Duke Health and Avanade, a joint venture between Microsoft and Accenture that is now majority owned by the latter, have the Smart AI Governance Engine to help health systems take stock of the AI they’re using.

AI chipmaker Nvidia and Aidoc, a medical AI company, announced they are working on a set of standards to help AI companies understand how to build products that take into consideration how doctors work.

“The point of this is, can we offer something by codifying our knowledge and understanding?” said Kimberly Powell, vice president of health care at Nvidia, who sees its forthcoming blueprint as complementary to existing standards-setting groups.

A jumble of coalitions, including Microsoft’s Trustworthy and Responsible AI Network; Duke’s Health AI Partnership; VALID AI, which was originally developed out of the University of California Davis; and the Coalition for Health AI, whose members include the Mayo Clinic and Johns Hopkins, are all vying to develop standards for health AI.

Why it matters: While the Food and Drug Administration regulates AI diagnostic tools and AI within medical devices, not all health care AI falls within its jurisdiction. HHS’ Office of the National Coordinator for Health Information Technology is starting to regulate AI within electronic medical records and its Office for Civil Rights now prohibits health systems from using discriminatory algorithms.

Still, there are big gaps. No one regulates AI that takes down doctors’ notes or AI chatbots that interact with patients.

What’s next? HHS is expected to lay out its strategy for regulating AI in health care by January 2025. Assistant Secretary for Technology Policy Micky Tripathi and HHS Deputy Secretary Andrea Palm have said that quality assurance labs will be part of the plan.

WELCOME TO FUTURE PULSE

Edgewater, Md.

Edgewater, Md. | Shawn Zeller/POLITICO

This is where we explore the ideas and innovators shaping health care. 

Fossilized footprints made by a sparrow-size Tyrannosaurus rex relative in South Korea suggest the tiny feathered dinosaur flapped its wings to run faster , a possible precursor to flying, Science reports.

Share any thoughts, news, tips and feedback with Carmen Paun at cpaun@politico.com, Daniel Payne at dpayne@politico.com , Ruth Reader at rreader@politico.com, or Erin Schumaker at eschumaker@politico.com.

Send tips securely through SecureDrop, Signal, Telegram or WhatsApp. 

INNOVATORS

PERTH, AUSTRALIA - FEBRUARY 17: Kim Garth of Australia wipes away sweat during day three of the Women's Test Match between Australia and South Africa at WACA on February 17, 2024 in Perth, Australia. (Photo by Paul Kane/Getty Images)

Caltech researchers want to examine that sweat. | Getty Images

The Biden administration will help California Institute of Technology researchers develop a wearable sweat-sensing system to measure chronic pain and University of Iowa scientists figure out whether personalized nanoparticles and a woman’s own immune system can treat ovarian cancer.

Those are among the projects getting a slice of $110 million in funding from the Advanced Research Projects Agency for Health as part of its Sprint for Women’s Health, the agency announced.

The sprint is part of the White House Initiative on Women's Health Research that President Joe Biden announced in 2023. First lady Jill Biden touted the funding at a health care conference in Las Vegas today.

"ARPA-H is de-risking investment in these big ideas," she said, adding that women's health issues often go unaddressed. "You know the woman," she said. "The woman who goes to the doctor and leaves with more questions than answers."

Why it matters: Women have long been underrepresented in clinical trials. The downstream effect of not being included means that many drugs and treatments aren’t developed with women in mind and are less effective and safe for women.

Being underrepresented in research also makes it more likely that women will be misdiagnosed or go undiagnosed for some diseases and conditions.

By the numbers: ARPA-H selected its 23 awardees from more than 1,7000 submissions from 34 countries, 45 states and Washington.

Among the other grantees:

— the Children’s Research Institute in Washington will seek to develop an accessible real-time assessment of chronic pain based on women’s eyes response to light.

— Massachusetts General Hospital in Boston wants to build a non-invasive wearable headband for at-home use to detect precursors of Alzheimer’s disease.

— The Wyss Institute in Cambridge, Massachusetts, aims to devise an implantable lymphoid organ as a cancer therapy to treat late-stage and metastasized ovarian cancer.

THE LAB

Radiologist Fredric Pedulla shows a nodule on a lung scan of a smoking person for the Acapulco experimentation in Ajaccio on December 16, 2021 on the French Mediterranean island of Corsica. - Corsica wants to be at the forefront of lung cancer screening, the leading cause of cancer deaths in France and particularly active on the island, thanks to the launch of a study to detect "early lung lesions" by a   low-radiation scanner. (Photo by Pascal POCHARD-CASABIANCA / AFP) (Photo by PASCAL POCHARD-CASABIANCA/AFP via Getty Images)

AI capable of examining images didn't always perform better in diagnosing cancer cases, a study found. | AFP via Getty Images

More information doesn’t always mean a better bot.

In a study comparing 10 AI chatbots’ performance in evaluating cancer cases, Canadian researchers found that AI systems which could accept image inputs didn’t always perform better than those that could not.

Chatbots that could evaluate images were less accurate when reviewing cases with multiple images compared with those with a single image.

And the top performing chatbot on multiple-choice questions wasn’t in the top three of free response performers.

The ways questions were posed mattered. AI systems performed significantly better when asked multiple-choice questions compared to open-ended ones, the researchers found.

Why it matters: Evaluating the accuracy of AI systems in health care remains complex and often uncertain, creating challenges for clinicians looking for assurance that the systems work.

Some AI systems that perform well in testing may not in the real world.

Even so: The study was small, testing 10 chatbots on 79 cases.

 

Follow us on Twitter

Carmen Paun @carmenpaun

Daniel Payne @_daniel_payne

Ruth Reader @RuthReader

Erin Schumaker @erinlschumaker

 

Follow us

Follow us on Facebook Follow us on Twitter Follow us on Instagram Listen on Apple Podcast
 

To change your alert settings, please log in at https://login.politico.com/?redirect=https%3A%2F%2Fwww.politico.com/settings

This email was sent to edwardlorilla1986.paxforex@blogger.com by: POLITICO, LLC 1000 Wilson Blvd. Arlington, VA, 22209, USA

Unsubscribe | Privacy Policy | Terms of Service

No comments:

Post a Comment

PNP sues Sara for 'assault,' AFP revamps her security

The Quezon City Police District on Wednesday filed the first criminal complaint against Vice President Sara Duterte that stemmed from her ou...