A team of researchers at Arizona State University (ASU) has created a powerful new AI tool called Ark+. This tool helps doctors read chest X-rays more accurately, which can lead to better health care for patients.
“Ark+ is made to be open, reliable, and useful in real hospitals,” said Professor Jimmy Liang, who led the study published in the journal Nature.
In tests, Ark+ showed it could spot many different diseases in the lungs — even rare or new ones like COVID-19 and bird flu. It was more accurate than other tools made by big companies like Google and Microsoft.
Despite high health care spending, the U.S. ranks low in life expectancy, behind countries like Cuba and Qatar. Patients want better care, and doctors want to get diagnoses right the first time.
That’s where AI like Ark+ can help.
Our goal was to build a tool that not only performed well in our study but also can help democratize the technology to get it into the hands of potentially everyone. Ultimately, we want AI to help doctors save lives.” More bang for the health care buck.
Jianming “Jimmy” Liang, ASU professor from the College of Health Solutions
Why Chest X-rays?
Chest X-rays are common and useful. They help doctors quickly check for:
- Lung diseases like pneumonia or tuberculosis
- Heart problems
- Broken ribs
- Some stomach issues
But reading them can be hard — even for experienced doctors. Rare or new diseases may be missed, like what happened early in the COVID-19 pandemic.
What Makes Ark+ Special?
Ark+ helps by:
- Reducing errors
- Making diagnoses faster
- Offering free, high-quality AI tools to the world
“We believe in open science,” Liang said. “That’s why we used public and global data to train the AI.”
How It Works
Ark+ was trained on over 700,000 chest X-ray images from around the world. It also learned from doctor notes written for each image — something other AI tools often ignore.
“Those expert notes help the AI learn more,” said Liang. “It’s like learning directly from top doctors.”
Most big companies train AI using only images, without the expert notes. Liang’s team took a different approach by using those notes to make the AI smarter.
A Small Team With Big Results
Liang’s team was small — including graduate students — but had support from:
- The National Institutes of Health
- The National Science Foundation
- Mayo Clinic Arizona
Despite being small, they built a tool that beat some of the biggest names in tech.
Key Features of Ark+:
- Trained on global data: So it works well across different populations
- Open and shareable: Other researchers can use and improve it
- Learns fast: Can identify rare diseases with few examples
- Easy to adapt: Can be adjusted to find new lung problems
- Fair and secure: Works well even with uneven data and protects patient privacy
One big reason Ark+ is powerful: it’s free and open to all.
“If we try to compete directly with big companies, we may not win,” said Liang. “But by working with others, we can be stronger than any one company.”
Looking Ahead
Ark+ can also be used for other types of medical images like CT scans or MRIs. The team hopes it will become a base for future AI tools in health care.
They want hospitals everywhere — even in small towns — to benefit.
Their mission: to make AI in health care safer, smarter, and available to everyone.
“By sharing this tool openly,” Liang said, “we invite others to help us make medical AI more fair, accurate, and easy to use.”
And that could be great news for health care in the U.S. — and around the world.
Source: Arizona State University
Journal reference:
Ma, D., et al. (2025). A fully open AI foundation model applied to chest radiography. Nature. doi.org/10.1038/s41586-025-09079-8.
Leave a Reply