New CT technology reduces radiation dose by 1/4

New CT technology reduces radiation dose by 1/4. August. 19, 2016 07:16. by 변지민 동아사이언스기자here@donga.com.

A Korean research team has discovered a new method of reducing one fourth of radiation dose exposed by computed tomography (CT). The team led by Dr. Ye Jong-cheol, an endowed chair professor of the Bio and Brain Engineering Department at KAIST, announced on Thursday that they developed a new artificial intelligence technology that improves quality of medical images filmed by low-dose rate CT.

When CT is used in medical examinations to obtain tomographic images inside a patient’s body, the tomography exposes high level of radiation. Moreover, the “low-dose rate CT,” which cuts radiation, still lacks in definition, as it is disturbed at even small noises due to weak signal.

A “filtering” technology is required to detect and eliminate noises, but it was not easy to distinguish image signals and interfering ones, which are usually jumbled up.

To solve this problem, the team applied the “machine learning” approach; a common trait found in artificial intelligence systems. By segmenting signal components into various directions, the team programmed an algorithm to let the artificial intelligence find out where the actual noise came from. Based on this approach, the image filmed at 2.5mSV (millisievert) low-dose rate CT was converted into high-definition image, which showed similar clear images filmed at 10mSv with a high-dose rate CT. On average, Korean people receive a radiation dose of less than 3.6mSv.

“This new technology, which can be commercialized anytime soon, can significantly reduce radiation exposure levels associated with CT,” Dr. Ye said.

Thanks to the new development, Dr. Ye has been ranked second at the “Low Dose CT Grand Challenge” held by the Annual Meeting of the American Association of Physicists in Medicine (AAPM) held on Aug. 1. The six-month online competition started in December 2015 drew 103 teams worldwide with an aim to reduce CT radiation exposure. The findings will be published in the academic journal “Medical Physics.”