A student in the US has developed a new automated X-ray machine that can cut the time it takes to examine 500 microscopic particles of air pollution from 44 hours to just 10 hours.
The technology, known as X-AutoMap, was developed by Carlos DeLeon during his internship at the National Synchrotron Light Source II at Brookhaven National Laboratory (BNL). DeLeon participated in the U.S. Department of Energy (DOE) Science Undergraduate Laboratory Internship (SULI) program.
Under the supervision of Ajith Pattammattel, PhD, a beamline scientist in the lab, DeLeon worked to automate the X-ray analysis process. The tool allows an X-ray microscope to automatically identify particles of interest and examine their chemical composition in more detail.
The software could make it easier to understand how air pollution affects human health. It could also be useful for studying batteries and electronics. “It would be much more efficient if we could determine detailed chemical compositions without a human having to control the experiment,” said Pattammattel.
Automated X-ray scans
Air pollution contains thousands of microscopic particles with different chemical compositions. This can include soot, dust, pollen and microplastics in the air that people breathe in every day. Some pose greater health risks depending on their chemical composition.
However, identifying them can take several days as researchers have to manually examine large samples and select individual particles for more detailed analysis. To speed up the process, the team developed X-AutoMap software to work with the facility’s Hard X-ray Nanoprobe (HXN).
Photo credit: Timothy Kuhn / Brookhaven National Laboratory
The HXN is a powerful microscope that can examine materials at extremely small scales. It first performs a quick scan of a sample at low resolution. The software then analyzes X-ray fluorescence signals to detect particles that match specific chemical patterns.
When the software finds the particles, the microscope automatically starts a high-resolution scan to display their chemical composition and structure. Pattammattel explained that the software follows scientific rules that the researchers have established in advance.
He explained the reason for this approach. “We thought it would be much easier if we could encode different pieces of expert knowledge and combine them into a computer model so that the microscope could use this information to automatically identify where high-resolution scans would be needed,” Pattammattel added.
Shortening the exam time
In laboratory tests, the scientists used fine dust samples that they collected from urban air pollution. The software successfully examined around 500 particles in 10 hours. In contrast, manual data collection took a total of 44 hours.
The system also revealed key differences between two types of iron-containing particles. Some particles contained iron, chromium and manganese, densely mixed. Many had a spherical shape associated with high-temperature industrial processes.
Others contained iron, calcium and silicon arranged in groups of individual grains. This structure is often associated with natural or geological materials. According to the team, these differences can be difficult to detect in lower resolution scans.
But because the software can only identify particles specified by scientists and may miss very small ones or those with weak chemical signals, the team plans to improve it and explore AI models to detect more materials.
DeLeon reflected on his internship experience and encouraged hesitant students to apply for similar opportunities. “So many people tend to doubt themselves or sell themselves short and miss opportunities like this,” he concluded in a press release. “You really never know where an experience like this can lead, so I encourage anyone interested to apply and see what happens.”
The study was published in the scientific journal Advanced intelligent detection.
