Artificial Intelligence in Healthcare: More Than Just Opportunities.

2020-01-02

Artificial intelligence has enormous potential in the field of healthcare, improving diagnoses and finding new, more effective drugs. However, as discussed in a recent article in Scientific American, the entry of artificial intelligence into the medical field also brings many new challenges and risks. In the past five years, the US Food and Drug Administration has approved over 40 different artificial intelligence products. However, according to Scientific American, none of the products sold in the US have undergone performance evaluations through randomized controlled clinical trials. If properly tested and controlled, artificial intelligence systems can become powerful tools, helping healthcare professionals identify overlooked symptoms and improve health outcomes. For example, an artificial intelligence tool used to detect diabetes and eye diseases was tested on hundreds of patients, and the results showed that it helped identify early signs of diabetes and retinal diseases. If suspicious symptoms were detected, patients were referred to medical experts. Without careful testing, artificial intelligence systems used to assist in diagnosis and treatment can potentially harm patients instead of preventing harm. The Scientific American article details a potential issue with relying on artificial intelligence to diagnose diseases, using an example of artificial intelligence analyzing chest X-rays to detect patients who may have pneumonia. While the system proved accurate during testing at Mount Sinai Hospital in New York, it failed in image testing at other hospitals. Researchers found that the artificial intelligence was able to distinguish between images generated by portable X-ray systems and those generated by radiology departments. Using portable chest X-ray systems on patients who are unable to leave their beds poses a greater risk of misdiagnosis. Artificial intelligence systems are typically trained on electronic health records, but many of these records are often incomplete or incorrect because their primary purpose is usually billing rather than patient care. For example, a survey conducted by KHN found that patients' medical records sometimes contain life-threatening errors, such as medication lists that include inappropriate drugs. In addition, diseases are often more complex than artificial intelligence engineers and scientists can anticipate. Therefore, as artificial intelligence becomes more prevalent, it is crucial for developers to collaborate with healthcare agencies to ensure thorough testing of their systems. Regulatory agencies also need to establish standards for the reliability of artificial intelligence diagnostic tools.

RECENT AI NEWS

RECENT AI TOOLS