Whether it's an online commercial database or a home-grown system, data mining is informing--and transforming--how clinicians treat patients. But although such systems can process vast amounts of structured data, they still have trouble with more nuanced pieces of information, notes an article at Kaiser Health News.
Researchers and tech companies are working on systems that can access unstructured text from doctor's notes or an individual patients' full range of symptoms and treatments and incorporate this information into its recommendations.
Providers are already mining text in published materials. For example, Baylor College of Medicine launched a project to mine data from research papers to improve cancer treatment and a tool created with IBM's Watson technology pulls from reference materials, clinical guidelines and medical journals in real time to help doctors diagnose patients and solve medical problems.
But, Peter Szolovits, director of MIT's Clinical Decision Making Group, told KHN, computers are "notoriously bad" at understanding English.
Developers also have an eye on decision-support tools that can understand and learn, according to the article, just as Google uses artificial intelligence to deliver tailored search results.
Watson, perhaps, is the best-known example of the expanding use of artificial intelligence--a venture that Big Blue is expanding. It recently paired with the New York Genome Center in an effort to bring more personalized treatment to cancer patients, for example.
The expense and difficulty of gaining high-quality datasets for these ventures remains one of the challenges, as well as developing smart models and teaching them to detect patterns, the KHN article notes.
To learn more:
- read the article