A primer on the language of artificial intelligence

AI is coming to a practice near you (if it isn’t there already). This primer helps explain NLP in healthcare and how different types of AI interconnect.
NLP in healthcare

The healthcare industry stands to gain a great deal by capitalizing on the current wave of AI advancements. With healthcare organizations predicted to spend up to 11% of their budgets on the technology in 2024, it has become an imperative that providers gain a basic understanding of how AI works. For those without a solid background in computer science, this rapid acceleration into adoption can be daunting. But simply believing the AI hype isn’t the answer and would leave those with less knowledge of data science ill-prepared to assess the actual risks and challenges inherent in any new technology. Below we discuss some key terms and definitions that providers should be aware of as AI becomes a bigger presence in their practices.

To begin, let’s establish a baseline on what we mean by artificial intelligence (AI). Simply defined, AI refers to the ability for computer applications to reason, learn from, and/or make decisions about information in a manner similar to humans. In essence, AI takes incoming information, processes it, and outputs predictions or results – all in a way similar to how the human brain works. (Interestingly, despite the attention it has received lately, AI has been around since at least the 1950s, although sources vary as to when the first official AI application was developed).

Evolving applications of AI

Healthcare has relied on AI tools since the 1970s, when the first algorithm for detecting blood infection treatments was developed. For example, clinical decision support (CDS) algorithms use a type of AI called rules-based expert systems which use a specific set of human-supplied rules to guide decision-making. Another broad type of AI called machine learning (ML) allows systems to recognize patterns in incoming information or use trial-and-error experimentation to infer results. For example, ML drives detection algorithms used in many medical devices such as automated electronic defibrillators (AEDs) and smart monitors.

Another more complex form of AI is deep learning, a type of ML, which uses neural networks. These neural networks contain several layers of algorithm processing nodes that function like the neurons in a human brain. This method can “learn” as it receives more information, ultimately leading to more accurate output. The research into the application of these neural networks to healthcare is still ongoing, but so far, many potential use cases have been put forward. Neural networks have demonstrated the ability to detect and diagnose anomalies in medical images, identify cohorts for clinical studies from aggregated patient data, predict patient no-shows, draft responses to patient inquiries, and much more.

Deep learning is able to perform such a multitude of complex tasks because it is built upon a data model that was determined by human experts. Essentially, a data model is what the AI uses as a reference to make inferences when presented with novel information. Large language models (LLM), for example, are models of semantic and syntactic rules that govern a particular language. LLMs can be used to perform several tasks such as natural language processing (NLP).

NLP in healthcare unlocks unstructured data

NLP – or the ability for computers to interpret and manipulate human language – has seen a recent rise in interest in the healthcare industry. This is due, in part, to its potential to simplify many time-consuming administrative and documentation tasks – from speech-to-text dictation and computer-assisted-coding to enhanced CDS and data mining. Perhaps one of the most cutting-edge applications of NLP is the processing of unstructured data – such as clinical notes – which contains important patient information not otherwise captured in structured entries. NLP can help unlock the information from these unstructured sources, and, by training the underlying LLM on a robust clinical terminology, properly integrate relevant information into the EHR. This in turn saves providers time that would otherwise be spent transcribing these notes manually or having to read through the entire note when searching for specific information.

While we have had AI in healthcare for a long time, recent developments in the field have attracted intense interest from key decision makers in the industry. Healthcare organizations are pivoting to invest a large amount of money into the latest AI applications, while regulators are rushing to develop guidelines for its appropriate use. Yet if we take a step back and familiarize ourselves with the basics of AI, healthcare providers can be better positioned to make well-informed decisions on how to make the most of the potential of AI while shielding against some very real risks. 

For more on AI, NLP, and LLMs in healthcare, download our insight brief, How to train your LLM.

Ideas are meant for sharing.

Sign up today and have Ideas delivered straight to your inbox.

Latest Ideas​

Article
Discover how clinical terminology serves as the Rosetta Stone of health IT, influencing everything from documentation to billing and beyond.
Article
Learn how incorporating LLMs into value set creation optimizes clinical workflows, boosts accuracy, and enables scalable, clinically relevant management.
Article
From helping NASA astronauts to payers, providers, and health tech, IMO Health has achieved a lot throughout its history – and we’re...