一系列文档和参考实现，重点介绍如何将各种技术组合成有意义的医疗保健模式-Java-Shell-下载

# Enrichment # Table of Contents - [Welcome to health-patterns](#enrich-overview) - [How to deploy](#how-to-deploy) - [Using the pattern](#using-the-pattern) - [Options](#options) - [Advanced topics](#advanced-topics) ## Welcome to health-patterns health-patterns is a place to find cloud reference implementations for the overall Alvearie architecture (https://alvearie.io/architecture) that incorporate best practices using open technologies. Each pattern incorporates parts of [Alvearie](https://alvearie.io/) along with other open technologies that can be used to start building your own healthcare solutions using a common base of proven technology. #### Enrichment Pattern Overview The **Enrichment** health pattern is a reference implementation of a clinical data enrichment process. It allows medical data to be enriched using various components, such as de-identification, terminology normalization, and the [ASCVD](https://github.com/LinuxForHealth/health-analytics/tree/main/ascvd) analytic evaluation. The Enrichment Flow is designed to read medical data (in FHIR format) from a configured [Kafka](https://kafka.apache.org) topic. As the data is processed, any errors that are detected are logged and posted to target kafka topics (if configured). Once complete, the updated FHIR data is posted back to a Kafka topic for further use. The current flow is designed to operate on FHIR resources only. Currently, for any FHIR data that enters the pipeline, we support the following enrichment steps: - NLP Insights: This step will process FHIR resources (or a bundle of resources). It will apply natural language processing to certain resource types in order to enhance resources by adding additional information to the resource or by adding new resources based on text found in the original resource. At present, the supported resource types are AllergyIntolerance, Immunization, DiagnosticReport, and DocumentReference. For configuration details, see the [nlp-insights documentation](https://github.com/LinuxForHealth/nlp-insights). - FHIR Terminology Service: This step will update the clinical data by adding/updating values to adhere to the terminology service configuration. The configuration mapping rules are currently static and can be found [here](https://github.com/LinuxForHealth/health-patterns/tree/main/services/term-services-prep/src/main/resources/defaultMappings) - [De-Identification](https://github.com/Alvearie/de-identification) Service: This step will de-identify the clinical data flowing through the pipeline and store the de-identified version in a separate FHIR server. The de-identification rules are currently static and can be found here. - [Million Hearts ASCVD Model](https://github.com/LinuxForHealth/health-analytics/tree/main/ascvd): This step will calculate a ten-year risk of cardiovascular disease using the Million Hearts ASCVD Model. ## How to deploy Note: Although the Enrichment pattern can be deployed on its own by following these steps, it will also be deployed as part of the default Ingestion pattern deployment [see Ingestion](../ingest/README.md). #### Prerequisites These instructions assume that you have the following resources, tools, and configurations: - Kubernetes cluster 1.10+ - Helm 3.0.0+ - Access to ```kubectl```, the Kubernetes command line tool - PV provisioner support in the underlying infrastructure **Note:** If you don't have access to a kubernetes cluster, this pattern can also be [deployed to minikube](../helm-charts/health-patterns/README_minikube.md). #### Check out the code ``` git clone https://github.com/LinuxForHealth/health-patterns.git cd health-patterns/helm-charts/health-patterns helm dependency update ``` Note: by changing the directory as shown above, you will be in the right place for future file access. #### Create a new namespace Please note that although this is step is optional, it is highly recommended that you create a new namespace in your Kubernetes cluster before installing the pattern. This will help prevent the various artifacts it will install from mixing with other artifacts that might already be present in your Kubernetes cluster. To create a new namespace called ```your-namespace``` and make it your default for future commands: ```bash kubectl create namespace your-namespace kubectl config set-context --current --namespace=your-namespace ``` **NOTE:** The length of a namespace name must be less than or equal to **20 characters**. Using a name that is longer than 20 characters will result in a failure to deploy the Nifi pod due to a certificate issue (the error will be visible in the NifiKop log). #### Update the internalHostName The internal host name used to communicate with nifi requires that you substitute your custom namespace (created above) into the value shown below. For example, if you created a namespace called `your-namespace` then the update in the `values.yaml` file would be ``` # Update "alvearie.svc" to "<your namespace>.svc" internalHostName: &internalHostName alvearie-nifi-0.alvearie-nifi-headless.your-namespace.svc.cluster.local ``` #### Set the releaseName To use the enrichment pattern, you should change the release name to `enrich` by default. The `values.yaml` file contains a `releaseName` variable. ``` releaseName: &releaseName enrich ``` #### Disable Ingestion Flow By default, both the ingestion flow and the enrichment flow are configured for installation. If you are only deploying the enrichment pattern, then you can disable ingestion by changing the `values.yaml` file entry for ingestion to be `false`. ``` ingestion: enabled: &ingestionEnabled "false" enrichment: enabled: &enrichmentEnabled "true" ``` #### Ingress parameters We recommend exposing the services in this chart via ingress. This provides the most robust and secure approach. If you choose to expose services via port-forwarding, load-balancer, or other options, please be careful to ensure proper security. Ingress requires a specific ingress class to be used. Different cloud providers rely on different ingress classes, so choose the one that matches your cloud provider. For example, some possible choices might be: - IBM: public-iks-k8s-nginx - Azure: addon-http-application-routing - AWS: nginx You will also need to provide a hostname for your ingress. What this is and how it gets created will be unique to your cloud infrastructure. Once you know these values, use both of them to update the `ingress` section of the file ```helm-charts/health-patterns/values.yaml``` as shown below. Note that the ingress class currently defaults to `public-iks-k8s-nginx` so if that is your choice, no update to the ingress class is needed. However, the ingress hostname **MUST** be updated. ``` ingress: enabled: &ingressEnabled true class: &ingressClass public-iks-k8s-nginx hostname: &hostname replace-me ``` For example, to deploy in the IBM Cloud environment, we would add ``` ingress: enabled: &ingressEnabled true class: &ingressClass public-iks-k8s-nginx hostname: &hostname <<your-ibm-hostname>> ``` #### Deployment The following Helm command will deploy the enrichment pattern. The enrichment pipeline will be ready to accept FHIR data and run it through a series of steps producing a possibly updated FHIR bundle with resources that have been added or modified by the enrichment process. ``` helm install enrich . ``` After running the command above, you will see notes that give you information about the deployment, in particular, where the important services (e.g. Nifi, expose-kafka) have been deployed. **IMPORTANT NOTE** The release name for the enrichment pipeline must be **enrich** (see [Advanced topics](#advanced-topics) for additional information) #### Uninstall/delete To uninstall/delete the deployment, use: ``` helm delete enrich ``` Deletion of charts doesn't cascade to deleting associated `PersistedVolume`s and `PersistedVolumeClaims`s