Machine learning can help improve the overall performance of CT scans, reducing radiation exposure and boosting image quality, according to new findings published in Nature Machine Intelligence.
The research comes from engineers at the Rensselaer Polytechnic Institute in Troy, New York, and a team of radiologists at Massachusetts General Hospital and Harvard Medical School, both in Boston.
“Radiation dose has been a significant issue for patients undergoing CT scans,” Ge Wang, the study’s corresponding author and a professor of biomedical engineering at Rensselaer, said in a news release from the institute. “Our machine learning technique is superior, or, at the very least, comparable, to the iterative techniques used in this study for enabling low-radiation dose CT. It’s a high-level conclusion that carries a powerful message. It’s time for machine learning to rapidly take off and, hopefully, take over.”
The study’s authors developed a dedicated deep neural network for low-dose CT image reconstruction, finding that their approach “performs either favorably or comparably in terms of noise suppression and structural fidelity, and is much faster than commercial iterative reconstruction algorithms.” The network was also faster than other existing techniques.
“We are excited to show the community that machine learning methods are potentially better than the traditional methods,” Wang said. “It sends the scientific community a strong signal. We should go for machine learning.”
“Professor Wang’s work is an excellent example of how advances in artificial intelligence, and machine and deep learning, can improve biomedical tools and practices by addressing hard problems—in this case helping to provide high-quality CT images using a lower radiation dose,” Deepak Vashishth, director of the Center for Biotechnology and Interdisciplinary Studies at Rensselaer, said in the same release. “Transformative developments from these collaborative teams will lead to more precise and personalized medicine.”