National Clinical Research Center for Cancer in China signs agreement to use AI for breast cancer diagnosis
The aim is to improve diagnosis accuracy and encourage breast cancer screening, especially in rural and remote areas. With early detection, over 90% of cases are curable.
Xinhua, the official press agency of the People’s Republic of China (PRC), reported that the National Clinical Research Center for Cancer (NCRCC) in China has signed a deal with the Institute of Computing Technology under the Chinese Academy of Sciences (CAS) to use artificial intelligence (AI) in medical imaging.
The Tianjin Medical University Cancer Institute and Hospital was appointed as the NCRCC by the Ministry of Science and Technology of the PRC in 2013. CAS is the national scientific think tank and academic governing body in the PRC and it comprises 104 research institutes, 12 branch academies, three universities and 11 supporting organizations in 23 provincial-level areas throughout the country.
The first area of focus for the cooperation will be to use the AI technology for reading breast scans and mammograms, two common methods used in breast cancer screening. Deep learning technology will be used to build models based on the experience of radiologists. The machine would be trained on hundreds of thousands of breast scan reports before it assists the doctors in diagnosis. It will be able to produce a highly-accurate report in just a few seconds based on its reading of the scans.
Breast cancer is the most common malignant tumor among Chinese women, with about 272,000 new cases reported in China in 2015. More than 71,000 people died of breast cancer that year.NCRCC director Hao Xishan said that if found early, 95 percent of breast cancer patients could be cured. However, in rural and remote areas, patients often do not get tested until it is too late.
The aim of this agreement is to improve diagnosis accuracy and encourage breast cancer screening in regions of high prevalence and in rural areas, where experienced medical professionals are in short supply.
It has not been decided yet when the clinical use will begin and how many hospitals the technology will be implemented in.
AI has huge potential applications in the field of healthcare and radiology is one of the most promising areas within health. Learning algorithms are being used to read X-rays, CT scans, MRIs of all sorts.
As this New Yorker article explains, the initial use of computers for diagnosis was based on sets of rules. The systems did not improve with the the number of scans or images it was exposed to. But with machine learning, the machines are not programmed with rules. Rather they are provided with hundreds of thousands of images and their confirmed diagnoses. Ingesting these massive volumes of data, the machine trains itself to distinguish between say malignant and benign lesions. The machines are not expected to entirely replace radiologists. Rather they would play an augmentative role, boosting efficiency and accuracy.
They also have a limitation, called the ‘black box’ problem, as in they can provide a diagnosis but cannot explain why or how it arrived at the diagnosis. That is a problem common to deep learning based algorithms. Also, there are legal implications for liabilities and ramifications for the training of doctors (if they are no longer called upon to use their diagnostic skills on a regular basis, will they deteriorate?) which are yet to be explored. However, the increasing use of AI for taking over roles in healthcare which involve processing big volumes of data, is already well on its way.
Read the report on Xihuanet here.
Featured image from Wikimedia Commons ( work of National Institutes of Health, part of the United States Department of Health and Human Services)