Artificial intelligence (AI) has the potential to revolutionize healthcare, but integrating AI-based techniques into routine medical practice has proven to be a significant challenge. A plenary session at the virtual 2020 AACC Annual Scientific Meeting & Clinical Lab Expo will explore how one clinical lab overcame this challenge to implement a machine learning-based test, while a second session will take a big picture look at what machine learning is and how it could transform medicine.
Machine learning is a type of AI that uses statistics to find patterns in massive amounts of data. It could launch healthcare into a new era by mining medical data to find cures for diseases, identify vulnerable patients before they become ill, and better personalize testing and treatments. In spite of this technology’s promise, though, the medical community continues to grapple with numerous barriers to adoption, and in the field of laboratory medicine in particular, very few machine learning tests are currently offered as part of regular care.
A 10-year machine learning project undertaken by Ulysses G.J. Balis, MD, and his colleagues at the University of Michigan in Ann Arbor could help to change this by providing a blueprint for other healthcare institutions looking to harness AI. As Dr. Balis will discuss in his plenary session, his institute developed and implemented a machine learning test called ThioMon to guide treatment of inflammatory bowel disease (IBD) with azathioprine. With an approximate cost of only $20 a month, azathioprine is much cheaper than other IBD medications (which can cost thousands of dollars a month), but its dosage needs to be finetuned for each patient, making it difficult to prescribe. ThioMon solves this issue by analyzing a patient’s routine lab test results to determine if a particular dose of azathioprine is working or not.
Balis’s team found that the test performs just as well as a colonoscopy, which is the current gold standard for assessing IBD patient response to medication. Even more exciting is that clinical labs could use ThioMon’s general approach—analyzing routine lab test results with machine learning algorithms—to solve any number of other patient care challenges.
“There are dozens, if not hundreds of additional diagnoses that we can extract from the routine lab values that we’ve been generating for decades,” said Dr. Balis. “This lab data is, in essence, a gold mine, and the development of these machine learning tools marks the start of a new gold rush.”
One of the additional conditions that this machine learning approach can diagnose is, in fact, COVID-19. In the session, “How Clinical Laboratory Data Is Impacting the Future of Healthcare?” Jonathan Chen, MD, PhD, of Stanford University, and Christopher McCudden, PhD, of the Eastern Ontario Regional Laboratory Association, will touch on a new machine learning test that analyzes routine lab test results to determine if patients have COVID-19 even before their SARS-CoV-2 test results come back. As COVID-19 cases in the U.S. reach record highs, this test could enable labs to diagnose COVID-19 patients quickly even if SARS-CoV-2 test supply shortages worsen or if SARS-CoV-2 test results become backlogged due to demand.
Beyond this, Drs. Chen and McCudden plan to give a bird’s eye view of what machine learning is, how it works, and how it can improve efficiency, reduce costs, and improve patient outcomes—particularly by democratizing patient access to medical expertise.
“Medical expertise is the scarcest resource in the healthcare system,” said Dr. Chen, “and computational, automated tools will allow us to reach the tens of millions of people in the U.S.—and the billions of people worldwide—who currently don’t have access to it.”