Over the past 10 years, there has been a remarkable expansion in the use of deep learning (DL) systems in healthcare. Although artificial intelligence (AI) had its origins well over 50 years ago, we are now at a time in human history that may well be considered the “AI Renaissance.” This has come into being because of a perfect storm resulting from advances in technology and computer science as well as adoption of smartphones and the ability to extract data from a variety of resources.
Firstly, graphic processing units (GPUs), used to power high-resolution video games, have improved significantly over the last decade to the point where they include hundreds of cores and can perform millions of computations per second. Cluster GPUs in servers or place them in the cloud and you have an incredible amount of computer power that can be leveraged to perform DL analysis efficiently and inexpensively.
In addition, much information has become digitized, and paper records are becoming obsolete. Google and Apple enable the collection of information about where we go, what we buy, what we view, and much more. We now have healthcare sensors and connected devices that communicate remotely via the Internet. Because data is digitized, information exists in expansive labeled datasets that facilitate training of DL systems.
Lastly, there have been advances in “image detection systems” such that computers can analyze images—for example, X-rays—as well as advances in “natural language processing” so that DL programs can scrub and extract data from organized databases as electronic health records (EHRs). As a consequence, DL programs are now used extensively in finance, government, education, and business, as well as in healthcare.