Artificial intelligence (AI) is an umbrella term for any technology that simulates human intelligence (i.e. is “smart”). AI covers a wide range of technologies from facial recognition on social media to rovers for space exploration. The AI technologies currently most applicable for solving questions and problems in healthcare are rule-based expert systems, machine learning techniques, and physical robots or robotic processes.
Rule-based expert systems are from the first generation of AI, which came about in the 1970s. Their operations are relatively simple, making decisions based on “if-then” statements that are created by domain experts. One of the earliest successful expert systems was MYCIN, developed by Stanford to identify bacterial infections and suggest the appropriate antibiotics.
While technology has become much more sophisticated since then, medical diagnosis and treatment remains a target focus for AI. Many electronic health records come with an expert system today as well. Yet, expert systems often do not know how to deal with conflicting statements and changing the rules is time consuming, so expert systems is slowly being phased out by machine learning and big data.
Machine learning (ML) is one of the most common forms of AI and refers to a method of data analysis that is used to build models. ML models learn from a given set of data and identify patterns to make decisions on their own in a manner similar to humans. The majority of ML health applications utilize supervised learning, which requires an outcome variable of the data it reads in, such as whether or not a patient is diagnosed with a particular disease.
ML has spurred a whole field called precision medicine, which emphasizes the importance of personalized treatment approaches instead of one-size-fits-all approaches. Computers power precision medicine by evaluating a variety of attributes for each patient, like their genetics or lifestyle, to determine the best treatment approach for that specific individual. High costs and poor outcomes have long plagued healthcare, particularly in the United States, and AI can play a huge part in addressing gaps in healthcare.
Neural networks and deep learning are sophisticated forms of ML that also have tremendous potential to shape healthcare. Neural networks are weakly analogical to the functions of the brain. They are able to digest massive amounts of information to make predictions and for this reason, neural networks have long been used to analyze images. For instance, AI can be trained to identify lesions or other abnormalities to detect specific diseases with impressive accuracy.
Finally, robots powered by AI are another technology transforming healthcare, from automating tasks to directly helping with patient care. The iconic da Vinci Surgical Systems (check out the below video of the system performing surgery on a grape!) first came out in the early 2000s and demonstrated how powerful robotics can be applied to help solve healthcare challenges by improving accuracy of diagnosis and treatments.
AI in healthcare has become increasingly common, even extending to areas such as psychiatry and drug development. These technologies promise a new paradigm for healthcare and will soon become an inescapable part of patient care. The rest of this series will discuss more cutting-edge AI applications in healthcare, as well as the ethical questions surrounding their deployment.
Davenport, Thomas, and Ravi Kalakota. "The potential for artificial intelligence in healthcare." Future healthcare journal 6.2 (2019): 94.
Last Fact Checked on May 28th, 2021.