Artificial intelligence is assuming a greater role in many walks of life, with research suggesting it may even help doctors diagnose disease. One new study suggests artificial intelligence (AI) might someday detect breast cancer that has spread to the lymph nodes.
Researchers found that several computer algorithms outperformed a group of pathologists in analyzing lymph tissue from breast cancer patients. The technology was specifically better at catching small clusters of tumor cells—known as micrometastases.
Clinical pathologists examine samples of body tissue to help diagnose diseases and judge how serious or advanced they are. It’s painstaking work—and the hope, lead researcher Babak Ehteshami Bejnordi, of Radboud University Medical Center in the Netherlands, said, is that artificial intelligence can help pathologists become more efficient and accurate.
The study is the latest to delve into the idea of using artificial intelligence to improve medical diagnoses. Most of the algorithms in the study were “deep learning”-based, where the computer system essentially mimics the brain’s neural networks.
“To build the system,” Bejnordi explained, “the deep learning algorithm is exposed to a large dataset of labeled images, and it teaches itself to identify relevant objects.”
Dr. Jeffrey Golden is a pathologist at Brigham and Women’s Hospital in Boston. He agreed that artificial intelligence holds promise for “making pathologists more efficient.”
However, there’s a lot of work to be done before that is a reality, said Golden, who wrote an editorial published with the findings.
The study has its limits, he said. The computer-versus-human test was only a simulation exercise—and not truly reflective of the conditions that clinical pathologists work under.
So it’s not really clear how the algorithms would compare against pathologists in the workplace, Golden said.
That’s key because for any computer algorithm to work, there have to be digital images of tissue specimens to analyze. Cost and education—training pathologists in how to use the technology—are other issues, Golden pointed out.
For now, one thing seems certain: “Artificial intelligence will never replace the pathologist,” Golden said. “But it may improve their efficiency.”
The study tested 32 computer algorithms that were developed by different research teams for an international competition. The challenge was to create algorithms that could detect the spread of breast tumor cells to nearby lymph nodes, which is important in estimating a woman’s prognosis. The algorithms were tested against the performance of 11 pathologists, who independently analyzed 129 digitized images of patients’ lymph nodes. The doctors were given a time limit to accomplish the task.
In a separate test, the algorithms were pitted against one pathologist who was free of time constraints.
It turned out that some algorithms bested the pathologists who were under time limits. In particular, they outperformed humans when it came to detecting micrometastases.
Even the best-performing pathologist missed 37% of cases where the lymph tissue contained only micrometastases, the study found.
Ten of the computer algorithms performed better than that.
However, Golden said, the pathologists were facing obstacles they would not face in the real world.