Artificial intelligence can help with cancer diagnoses and may soon be used as an experimental technology. This is what researcher Wouter Bulten of Radboudumc writes in his dissertation.
According to Bulten, artificial intelligence can be used as an assistant to a pathologist, who investigates whether a patient has a tumor. “There is a global shortage of pathologists, and smart algorithms can help ease the pressure on them.”
This should lead to lower healthcare costs and faster diagnoses. Computer systems could also be used in population screening, after which pathologists scrutinize the suspected cases.
Researchers at Radboudumc have already concluded that artificial intelligence can help, but now the research has gone a step further, says Bulten. “After that we can start doing trials.” This means that the technology is put into practice, and the results that it produces are looked at.
According to Bulten, it is not the intention to completely sideline the pathologist, and a human view is still necessary.
Together with Google and the Swedish university Karolinska Institutet, Bulten organized a worldwide programming competition, in which approximately 1,250 programmers started working with research data. This meant that the software had to make diagnoses for prostate cancer.
“Such a competition is a good way to look for new techniques,” says Bulten. “As a researcher, you can try to solve the problem yourself, but if you ask thousands of others, it can lead to whole new creative ideas.”
A risk of using algorithms is that people allow themselves to be influenced by their judgment, even if they actually disagree with it. And that is precisely not the intention. “That is why we are looking at whether the algorithm can also indicate how certain it is of its case,” says Bulten.