NLP Insights

A Rest service for updating bundles of FHIR resources with discovered insights. The service is implemented as a Flask API within a docker container.

Purpose

The primary purpose of the discover insights API is to accept a bundle of FHIR resources and to return an updated bundle that includes discovered insights.

• Resources in the bundle may have been enriched by adding additional codes.
• For example an AllergyIntolerance resource for a peanut allergy might have UMLS code C0559470 or SNOMED-CT code 91935009 added to it.
• New resources may have been derived from unstructured text (such as clinical notes) contained within the bundle's resources.
• For example a DiagnosticReport that says the patient had a myocardial infarction might result in a derived Condition resource being added to the bundle.

Supported NLP Engines

The nlp-insights service requires an NLP engine service to perform NLP related tasks. We support two NLP services.

• IBM's Annotator for Clinical Data (ACD) and
• Open-source QuickUMLS

Quick Start

Our tutorials describe how to setup and configure nlp-insights with a supported NLP service. They also provide extensive description of how resources are derived and enriched:

• Tutorial for using the nlp-insights service with QuickUMLS
• Tutorial for using the nlp-insights service with ACD

Running the service locally

A local instance of the service can be started either by:

• Pulling a container image from our repository.
• Cloning our GitHub repo and building an image from the source code.

The directions for both approaches can be found here.

Although discouraged, it is possible to run the service outside of a docker container.

Kubernetes

The nlp-insights service is designed to be part of a larger health-patterns ingestion and enrichment pipeline. Helm charts are included so that the service can be deployed to kubernetes. The deployed service can then be integrated into a pipeline.

More details on deployment and configuration in a k8s environment are discussed here

HTTP Endpoints

The HTTP APIs for the service are described here. These APIs allow you to:

• Define the connection to the NLP engine service(s),
• Select the default NLP engine that will be used for insight discovery
• Discover insights
• Override the default engine and use a different NLP engine for one or more resource types

Build

We use gradle for all build and test related tasks. The important features are documented here.
Continuous Integration and documentation publishing is done with GitHub Actions, which is documented here.

Contributing

We welcome contributions! Please look at our contributing guide for details on how to begin.

License