How to Use Text Analytics for Health

In the past few years, artificial intelligence has made major strides in the arena of text analytics. Text analytics, sometimes called text analysis, is the use of artificial intelligence to glean useful data from unstructured texts. For certain applications, a computer can now parse text far faster than a human being.

Perhaps nowhere has this been more impactful than in the arena of public health. Clinical documents, discharge summaries, and other electronic health records all fall into the category of unstructured text. AI can now scan these documents for valuable information, and provide that information to health providers.

Microsoft’s Azure cloud service has been at the bleeding edge of text analysis and big tech health solutions. So it should come as no surprise that they’ve released a powerful new tool, Text Analytics for Health. Here’s everything you need to know.

What is Text Analytics for Health?

Text Analytics for Health is one of several tools that makes up Microsoft’s Azure Cognitive Services for Language. All of these tools are designed for intelligently processing the written word, but each has its own specific applications.

Medical documents have their own vocabulary and structure, and Text Analytics for Health has been specifically trained to understand them. The AI is capable of many unique features, including:

  • It can understand modifiers. A medical language processing tool needs to be able to understand fine distinctions. “Stroke” and “heat stroke,” for example, are linguistically similar. But in a medical context, they’re completely different. A software that made this kind of error could have a disastrous impact on patient outcomes. Text Analytics for Health’s assertion detection model sorts statements into three categories: certainty, conditional, and association.
  • It supports named entity recognition. The AI can scan text for specific terms and assign labels relevant to particular categories. For instance, it can distinguish medication names, symptoms, and diagnoses.
  • It’s specialized for medical use. Text Analytics for Health compares terms and concepts to a database of concepts, including the Unified Medical Language System (UMLS). This helps the AI disambiguate when a term is being used in a medical context. For example, did a patient suffer an electrical shock, or are they in a physical state of shock.
  • It’s capable of associating related concepts. The AI utilizes a technique called relation extraction to identify relationships between separate pieces of text. A good example of this is abbreviations, which it can associate with the full term.

How to Use Text Analytics for Health

Using Text Analytics for Health requires a bit of technical know-how. But operation is straightforward if you’re comfortable with language processing software.

All you need to submit is raw text. There’s no need to put your data into a special format. Assuming your text is ready to input, there are three ways to use the software:

  • Language studio – This is a web-based platform that requires no code or hardware. It’s somewhat limited, but it’s a good way to see if Text Analytics for Health will meet your needs.
  • REST API or Client Library – The API allows you to integrate Text Analytics for Health with your own in-house apps and FHIR platforms. This is incredibly powerful, since you can have complete control over the software.
  • Docker container – Docker container is a special tool that allows you to run the software on your own hardware. You’ll need to install the tool on your system and enable Linux support.

Getting Started With Text Analytics for Health

To use Text Analytics for Health, you’ll first need to create an Azure subscription. It’s free to sign up, and a free account will unlock enough features to preview the software. If you like it, you can upgrade to the (S) pricing tier to access the Analyze feature.

Once your account is created, you’ll need to find the key and endpoint from your resource. Copy this information to a text file and keep it handy. The software will request it when you connect your app to the Text Analytics for Health API.

From there, it depends on what interface you prefer. These are your options, with links to the instructions:

  • C# – This requires the Visual Studio IDE.
  • Python – This requires Python 3.7 or later.
  • JavaScript – This requires Node.js v14 LTS or later.
  • Java – This requires the Java Development Kit (JDK) version 8 or later
  • REST API – This requires the latest version of cURL.

How Can We Use Cognitive Analytics in Healthcare?

Cognitive computing is based upon self-learning, artificially intelligent systems that use pattern recognition and natural language processing to understand the meaning behind text. Instead of being “told” in advance how to understand the world, the system “learns” naturally, just like a human.

This has incredible potential for healthcare. Imagine an AI system that could monitor a patient’s breathing, heart rate, and brain activity, and dispense an anesthetic accordingly. It would be far safer than a human anesthesiologist, and could save many lives.

We’re not quite there yet. But cognitive computing is already making major changes to how we diagnose and treat patients.

It’s designed to deal with unstructured text. That’s text that’s not organized in any particular way, as opposed to a form or chart, which is structured. This allows AI to read text almost as well as a human, only much faster. It’s a powerful tool for diagnosing patients and making clinical decisions.

What is Microsoft Cloud for Healthcare?

Microsoft Cloud for Healthcare is a separate platform that’s related to Azure. It’s designed to provide a cloud-based platform for medical recordkeeping and decision-making. It isn’t part of Azure, but it integrates with it, and it has a managed, open-source FHIR server. This is a great resource for small practices which can’t afford to operate their own servers. And with Azure integration, you can seamlessly integrate Text Analytics for Health.

Final Thoughts

Text Analytics for Health is an impressive AI tool for improving patient outcomes. It’s a highly technical software; you’ll need to be very tech-savvy or have your own IT department to use it. But if you’re able to harness its capabilities, you and your patients stand to benefit.