Men in Brazil, diagnosed with colorectal cancer, usually refuse the surgical removal of the colon, one of the recommended treatments. After surgery, patients are required to have a colostomy bag for the rest of their lives. This is a real concern in the body-conscious Brazilian culture.
This aversion to the surgery has forced Brazilian oncologists to offer alternative therapies including radiation and chemotherapy. Surprisingly to doctors, about 20 percent of patients respond to those other treatments. However, doctors are hesitant to embrace these alternatives because they can’t predict which patient will fall into that group.
Now scientists are trying to answer that question by using artificial intelligence. The use of AI in healthcare, which was one of the topics discussed at GE’s recent Minds + Machines conference in Berlin, is a fast-growing field. Scientists are using so-called “deep learning networks,” which weave together hundreds, if not thousands, of data points and process this data with multiple algorithms simultaneously, mimicking the human brain. When crossing the street, pedestrians take into account dozens of factors, including the number and speed of approaching cars, the condition of the pavement, fellow travelers and even the shoes they are wearing or what they are carrying. Deep learning has the potential to do the same thing – but with even more data points and at speeds unmatched by humans.
This concept has actually been around for decades. The first artificial neural network was created in 1954 at the Massachusetts Institute of Technology. But is wasn’t until recently that the computing power was sufficient for effective calculations in a neural network.
Scientists working on the Brazilian problem are asking the AI to search for possible patterns that can predict who might be a successful candidate for the alternative cancer treatments.
They are feeding millions of data points into the cloud, including decades of colorectal data collected by national registries, thousands of MRIs and CT scans, gene panels and biomarkers. The software then looks for patterns, connections and correlations with a speed and detail unmatched by humans.
“We’re looking for the unknown,” says Dr. Michael Dahlweid, chief medical officer of GE Healthcare Digital, who is working on the AI. “We want to figure out if there is something that is too subtle to be found by humans.”
AI may also eventually be used to diagnose disease. The concepts in development rely on data being entered into deep learning systems. In these instances, clinicians and data scientists work hand-in-hand, reviewing the algorithms and fine-tuning them to help the AI as it learns to call out areas that doctors need to pay special attention to when interpreting body scans. “The idea is that after a few attempts or a few hundred attempts, suddenly, the AI becomes a reliable support to the people making medical care decisions,” Dahlweid says. The system doesn’t make the diagnosis, but brings suspicious findings to the doctor’s attention, who then decides on the further treatment.
AI could be a game-changing advancement in healthcare. “Last year, 70 startups were founded dealing with medical AI,” Dahlweid says. “It’s predicted that by 2021, investment in healthcare AI will reach $6.6 billion.”
As AI becomes a more common tool in healthcare, medical schools will have to change how they train physicians to make sure they have the new capabilities, skill sets and methodologies to use AI effectively, Dahlweid says. “In the future, I believe these techniques will become like a stethoscope, like penicillin – normal things,” he says. “But we have to learn how to work with them.”
Disclaimer: This is a technology in development that represents ongoing research and development efforts. These technologies are not products and may never become products. They are not for sale, and have not been cleared or approved by the FDA for commercial availability
 Farley and Wesley A. Clark (1954) first used computational machines, then called “calculators,” to simulate a Hebbian network at MIT. — Mind as Machine: A History of Cognitive Science By Margaret Ann Boden