Artificial intelligence may be just as good at detecting the spread of breast cancer as a specialist.
That’s the finding of a study by researchers in the Netherlands.
The scientists were examining whether artificial intelligence (AI) in the form of computer algorithms could perform as well as a pathologist when detecting the spread of breast cancer to the lymph nodes in women with the disease.
The researchers said they were stunned by what they found.
“AI is increasingly being recognized as a major element of the healthcare landscape. We are now at a turning point where AI algorithms perform as well as or better than clinicians at specific tasks. But still, I did not expect such remarkable results at this early stage. We showed that state-of-the-art AI algorithms perform as well as or better than pathologists in detecting the spread of breast cancer to lymph nodes,” Babak Ehteshami Bejnordi, an author of the study, told Healthline.
How AI procedure works
Bejnordi and his colleagues from the Radboud University Medical Center in Nijmegen in the Netherlands first produced computer algorithms to detect the spread of breast cancer as part of an international challenge in 2016.
The computer algorithms analyze tissue slides of sentinel lymph nodes.
Those are the lymph nodes that are closest to a tumor and the first place cancer would be likely to spread.
In this study, the researchers compared the performance of the algorithms against the performance of 11 pathologists who participated in a simulation exercise.
They found that some of the algorithms were better than pathologists at detecting the spread of cancer in an exercise with time constraints.
Without time constraints, some algorithms were as good as a pathologist in detecting the spread of cancer.
Although the evaluations that took place in this study still need to be undertaken in a clinical setting to determine if the same results can be achieved, Bejnordi says the use of AI in pathology could take a lot of pressure off specialists.
“Detection of cancer metastases in lymph node tissue is a complex, tedious, and time-consuming task. Pathologists may easily miss small metastases during diagnosis. Diagnosis of certain types of metastases such as metastases originating from lobular carcinoma can be notoriously difficult and error-prone. AI systems, in contrast, do not get exhausted and always make the same objective interpretation and therefore can aid the pathologists in their decision-making,” he said.
Artificial intelligence in medicine
Artificial intelligence is present in many aspects of modern life.
Speech recognition, computer chess games, and autonomous driving cars are just some of the ways in which AI is used.
The use of AI in medicine has taken a while to catch on, but in the past few years the implementation of the technology has seen a rapid acceleration.
In an editorial that accompanied Bejnordi’s study, Dr. Jeffrey Alan Golden, chair of the Department of Pathology at Brigham and Women’s Hospital in Boston, notes that “in 2014, the acquisition of AI startups in healthcare was about $600 million. In 2021, it is anticipated to be $6.6 billion or a 40 percent compound annual growth rate.”
AI involves the science and engineering that enables intelligent computer systems to perform tasks that require human intelligence.
Put another way, AI helps machines think and learn.
Golden believes there are numerous opportunities for this technology in medicine.
“One of the reasons medicine is so attractive is that the discipline has collected so much information or data on patients that it is impossible for a single person to integrate all of it into his/her thinking. A computer will likely be able to do so and use the data more effectively in helping guide physicians and other healthcare workers in the future,” Golden told Healthline.
AI may assist in improving diagnostics, but Golden believes human doctors will never be replaced by such technology.
“Looking into the future, I do not see a scenario where computers replace human doctors. Instead, they will make them better, more efficient, and safer. I view AI as a tool in the tool chest that healthcare works will be able to use to improve diagnosis, prognosis, treatment stratification, and the definition of intermediate treatment measures. It will assist and improve our ability to improve healthcare. It will be able to do analyses not possible by physicians. However, other things it will not be able to do,” he told Healthline.
Bejnordi agrees that AI will likely never completely replace doctors but will work alongside them and improve the efficiency of human doctors. He also anticipates that incorporating AI in a clinical setting will streamline the workflow of practitioners.
“The introduction of AI will soon offer a paradigm shift in how clinicians work, offering a major opportunity to boost workflow efficiency while at the same time allowing for more accurate and definitive diagnoses,” he said.
“Robust evaluations” of AI technology, he says, will be necessary for clinicians to trust the use of such technology.
Dr. Michael Blum, director of the University of California San Francisco (UCSF) Center for Digital Health Innovation, says evaluations in a clinical setting are crucial for ensuring AI performs as intended.
“As with every new technology, it will take some time to determine the best uses in healthcare and to work out the kinks. As the algorithms evolve out of the development space, there will need to be rigorous clinical validation to ensure that they function as intended and do not create unintended consequences,” he said.
Bejnordi and his colleagues are hopeful the algorithms they have devised will perform well in clinical studies.
He believes it won’t be long until such technology is used around the world.
“What matters most is providing the best care for patients. If the results of our clinical evaluations show that using AI makes us become more accurate, efficient, and confident in our diagnoses, it becomes immoral not to utilize this technology in practice,” he said.