Robots helping in operating rooms and spotting disease symptoms is no longer a concept for the future: Machine learning in medicine is already here.
According to PC Magazine, machine learning (ML) is an example of artificial intelligence (AI). After a limited amount of programming, machines learn an identification skill from viewing a data set of examples. The more examples a machine sees, the more accurately it can make predictions.
The power of machine learning is already being harnessed in a number of fields, and its applications in the healthcare field are constantly expanding.
Learn how AI is being used in Women's Health Ultrasound in the Meet the Expert video.
Learn how AI is being used in Women's Health Ultrasound in the Meet the Expert video.
Examples of Machine Learning in Medicine
Simple applications for machine learning include voice and facial recognition, search engine recommendations and language processors. In healthcare, ML is most commonly used in robotic surgery and for interpreting images in radiologic and pathologic settings.
Google developed an ML tool that could recognize breast cancer cells on a slide with 89% accuracy, as compared to the 73% accuracy of a human pathologist. ML may one day even allow robotic surgery unaided by humans.
Where Artificial Intelligence and Machine Learning May Lead Us Next
The full scope of how predictive analytics in healthcare will change the way physicians practice has yet to be realized, but some of the most promising uses being studied are in drug discovery, customized medications and disease prevention.
Perhaps one of the most well-known research databases currently being assembled is the National Institutes of Health's All of Us Research Program, which aims to sequence the genomes of 1 million people. As this program grows, its data on disease symptoms, treatment responses and general health has the potential to inform a host of ML projects in healthcare research.
However, Nature Materials cautions that the large data sets needed to transform clinicians' diagnostic capabilities through ML are vulnerable to malicious attacks. This underscores the need for data security within practices and even global standards for data protection.
Applying Machine Learning Data in Healthcare
Machine learning in medicine is promising, but data is only useful once it is applied. As the New England Journal of Medicine points out, a data set, no matter how large, is useless until it is "analyzed, interpreted and acted on."
Gynecologists use data — and data-processing technology — every day. For example, an ultrasound machine can store protocols for specific exams, ensuring that the physician is performing the correct scans and obtaining the images needed for both diagnosis and billing. The machine is creating a bank of diagnostic data and collecting the same data on each patient undergoing an exam for a suspected diagnosis. ML technology may one day allow ultrasound machines to learn which scans are suspicious in real time, even alerting the sonographer during an exam.
While it is likely that predictive analytics will reduce clinicians' workloads, it is unlikely that AI will eliminate the need for physicians altogether. After all, a computer may be able to learn, but someone has to set it on the correct course and verify its suggested diagnoses. Additionally, we still need doctors to interact with patients, because sometimes the right data needs to be revealed in careful, sensitive conversation. Machine learning in medicine will continue to shape and change the field, but it will never replace the need for skilled and compassionate practitioners.
