AI: Personal Therapy Robots and Other Ways that Machine Learning Is Changing Health Care

Supercomputers could lead to super advances in science and medicine. Morris MacMatzen/Getty Images

Many people worry that robots and artificial intelligence will eliminate jobs, including those in health care. But a little AI might not be unwelcome to people tired of waiting for test results or other improvements to their average experience at the doctor's office. AI could speed up care, make medications more affordable and give patients more time with their doctors. Here are four ways machine learning is already changing health care.

Diagnose Alzheimer's Earlier

When it comes to managing the progression of Alzheimer's disease, early diagnosis is key. But spotting the nascent signs of this neurodegenerative condition is challenging, which means by the time a patient is diagnosed, he or she is already experiencing many of the debilitating symptoms, such as loss of memory and cognitive function.

Harnessing the power of Big Data could completely change this and make it possible to predict who develops the disease. That could mean a patient gets the opportunity to start therapies that delay symptoms or completely prevent Alzheimer's.

The researchers, who published their study in Neurobiology of Aging, employed machine learning to develop an algorithm using PET scans of brains to search for the smallest signs of amyloid plaques. When detected in the brain, these proteins are one of the hallmark signs of Alzheimer's and may start to build up years (or even decades) before the onset of dementia.

The scientists found their formula predicted disease onset with 84 percent accuracy. This innovation could also be a boon for clinical trials, allowing medical researchers to easily identify patients who may be appropriate study subjects for novel therapies developed to prevent further progression of the disease.

Page Dr. Freudbot

Therapy is often seen as a first-world luxury. But AI has the potential to bring psychological counseling to the masses. Already, it has been used for challenging and costly tasks such as counseling Syrian refugees, according to The New Yorker. A number of companies have been working on chatbot products that communicate with users much like a trained professional would, meaning it's possible for machines to learn one of the most fundamental qualities of humanity: empathy.

One example is Woebot, a phone app that costs between $6 to $12 a week. The phone app, developed through a randomized controlled trial on stressed-out students at Stanford University, harnesses the power of cognitive behavioral therapy. This approach to counseling helps someone become more aware of—and change—problematic and troublesome behaviors that may alter the person's mood.

Like a real, live therapist you might pay $100 per hour to visit, this robot shrink will learn more about you over time and become a better counselor as a result. Unlike your therapist, though, this chatbot is there for you 24/7 at the most critical moments, like the middle of the night, when you're tempted to drunk-text your ex.

Advance Genomic-Based Medicine

Since completing the map of the human genome in 2003, the medical and scientific community has been excited about using the power of Big Data to identify causes and treatments for both common and rare diseases. The problem is that this giant pile of data is, well, giant. But AI is turning genomics into information with real clinical application—what is often called precision medicine.

This approach is already changing cancer care. Machine learning is helping classify tumors by genetic types, allowing doctors to match a patient's unique disease to the most effective treatments. In June, the American Society of Clinical Oncology announced that the IBM Watson supercomputer was able to generate potential tumor treatment recommendations for 96 percent of patients and reduced the time it takes for screening patients for clinical trials by 78 percent.

Repurpose Old Drugs

Scores of new pharmaceutical drug compounds are developed each year, but very few of them are effective for the diseases and conditions they're supposed to treat. Some don't even make to clinical trials, which means many are laid to rest on a shelf in a lab and never picked up again.

AI could help digest data in a way that recognizes disease states and drug pathways and connects them in unexpected ways. This task is typically too arduous for humans to do alone. Bioinformatics is able to expedite this process, meaning old drugs get a new lease on life (and could save lives in the process).