Tokyo's Jikei University Hospital and DeepMind Technologies Ltd., a London-based research firm, announced Thursday that they will begin developing artificial intelligence (AI) technology that can detect breast cancer at an early stage. The machine-learning technology will analyze mammograms from nearly 30,000 Japanese women that were provided by the hospital.
Google-affiliated DeepMind previously developed AlphaGo, a computer program that defeated a world Go champion in 2016. It trained the program by feeding it countless images of stone arrangements on a Go board.
While doctors diagnose breast cancer with the help of mammograms and other tools, thousands of cases around the world are said to go undetected each year. Identifying breast cancer in Japanese women is believed to be harder because of differences in their breast density and anatomical structure.
The research firm began work using images from about 30,000 people in Britain, but concluded it also needed data from Japan. It will further develop the AI using mammograms collected by Jikei from 2007 to 2018. The data from about 30,000 women who received mammograms during this period will remain anonymous.
DeepMind will program the AI to review the 30,000 images, along with 3,500 images from magnetic resonance imaging (MRI) scans, so that it more accurately identifies cancer that might otherwise be missed.
It may take several years for the system to be put into practice, as it is first necessary to verify the accuracy of diagnoses. The joint project is scheduled to last five years.
Norio Nakata, Associate Prof. at Jikei University sees the coming system as a substitute for the lack of radiologists in Japan. "Compared to other countries, the numbers of radiologists is small. We can expect the quality of medical services to rise with this system," he said.
The number of deaths from breast cancer in Japan is increasing every year, with 14,000 people dying of the disease in 2016, according to the Health, Labor and Welfare Ministry.
Read more from The Japan News at https://japannews.yomiuri.co.jp/
