The topic of artificial intelligence (AI) is of great interest worldwide. As all industries undergo a digital transformation, they are finding that the ability to use data is both a competitive advantage and strategic imperative. Healthcare is certainly no exception. In this first part of a two-part blog series, I will discuss some of the technologies that make up artificial intelligence, and in the second post present five use cases for AI technology in healthcare.
Within healthcare, as in many industries today, we are seeing a massive digital transformation taking place. A key driver for this is a need to manage costs and improve access against a backdrop of an aging population with increasing prevalence of chronic disease.
Digital solutions like telemedicine, wearables, and remote sensors are opening the door to innovation around patient engagement and preventative care.
And, we’re racing towards a world of precision medicine where genomics and other ‘omics become standard aspects of routine medical care.
These transformations are leading to the creation of massive new data sets, which creates an opportunity for the use of artificial intelligence.
Before we go further, let me establish a working definition for artificial intelligence. At Intel, we have developed a taxonomy for artificial intelligence that includes both machine learning and reasoning technologies that enable machines to more closely mimic human capabilities like sensing, reasoning, and acting.
Within the area of machine learning, there are two common approaches. The first is classic machine learning that uses statistical techniques like decision trees, random forests, and support vector machines. The second is deep learning, which uses neural networks. A common use for machine learning is classification, which is a common element in predictive models.
Adjacent to machine learning are reasoning systems, both memory, and logic based, which are able to identify patterns across data sets using various techniques and build prescriptive solutions.
These technologies are not mutually exclusive. They each have strengths and weaknesses. Think of them as various tools to solve specific problems. Increasingly we see solutions being developed that involve several of these technologies to solve specific aspects of a broader business need.
The critical precursor for any of these technologies is data. This is why digital transformation in healthcare is so important. Adoption of digital technologies is creating massive new data sets, which presents an opportunity for AI.
A common question that is posed to me is, “how does an organization get started with AI?”
To answer this question, it’s important to put AI in context with other analytics techniques.
AI is an evolution from traditional analytics. Key capabilities like data governance and training on data-driven decision making set the stage for being able to deploy AI solutions. Critical to this evolution is a top-down commitment from management to use data in order to drive business and clinical processes.
In the analytics maturity model shown above, you can see that artificial intelligence starts to come into play when organizations transition from using data to analyze what happened in the past, to using data to predict what will happen in the future. Often this transition occurs through the use of classic machine learning to build predictive models from structured data. Naturally, as organizations get more comfortable with this approach, they will look to utilize new, unstructured data sources that open the door for deep learning.
The next step after building predictive models is prescriptive models using cognitive systems that combine the prediction with a business process to recommend a course or courses of action.