Skip to content

README.md

Latest commit

 

History

History
204 lines (158 loc) · 15.5 KB

README.md

File metadata and controls

204 lines (158 loc) · 15.5 KB

Prerequisites

To use this solution accelerator, you will need access to an Azure subscription. While not required, a prior understanding of Azure Synapse Analytics, Azure Machine Learning, Azure Logic Apps, Power Apps and Azure Kubernetes Service will be helpful.

For additional training and support, please see:

  1. Azure Synapse Analytics
  2. Azure Machine Learning
  3. Azure Logic Apps
  4. Power Apps
  5. Azure Kubernetes Service

Deployment Guide

Start by deploying the required resources to Azure. The button below will deploy Azure Synapse Analytics, Azure Machine Learning and its related resources, Azure Cosmos DB, Function App, Logic App, Speech Service, Translator and Azure Kubernetes Service.

Step 1. Download Files

Clone or download this repository and navigate to the project's root directory.

More information on Cloning a repository

Step 2: Get the Required Datasets

  1. We are using the data provided by UCI Machine Learning Repository. You will need the following datasets (download them):
    • diabetic_data.csv

Step 3. Upload Sample Dataset

Step 3.1 Upload Sample Dataset to Azure Data Lake Storage

Upload the following files from the /Analytics_Deployment/Data folder into the ADLS storage account. Create under the container raw a new folder DatasetDiabetes

  • diabetic_data.csv
  • Names.csv

Step 3.2 Upload Sample Dataset to Azure Cosmos DB

Upload the following files from the /Analytics_Deployment/Data folder into the Cosmos DB. Each file matches the name of the container.

  • AdmissionSource.json
  • Appointments.json
  • ColumnLookupValues.json
  • ColumnNameMap.json
  • DischargeDisposition.json
  • ICD9Code.json

You can use the Data migration tool or go to the Data Explorer of your Cosmos DB, select the container and click Upload Item

Step 4. Update storage account permisions

In order to perform the necessary actions in Synapse workspace, you will need to grant more access.

  1. Go to the Azure Data Lake Storage Account created above
  2. Go to the Access Control (IAM) > + Add > Add role assignment
  3. Now click the Role dropdown and select Storage Blob Data Contributor
    • Search for your username and add
  4. Click Save at the bottom

OPTIONAL - Step 4.1: Add IP Adress to the Synapse Workspace

  • NOTE: If you did not allow all connections during deployment of your resources, follow the steps below

Before you can upload any assets to the Synapse Workspace you will first need to add your IP address to the Synapse Workspace. Before you can upload any assets to the Synapse Workspace you will first need to add your IP address to the Synapse Workspace.

  1. Go to the Azure Synaspe resource you created in Step 1.
  2. Navigate to Firewalls under Security on the left hand side of the page.
  3. At the top click + Add client IP client IP
  4. Your IP address should now be visable in the IP list.

Step 5. Add Compute to Azure Machine Learning

  1. Go to the Machine Learning Service and click the button Launch Studio
  2. Click on Compute - select Inference Clusters and click New
  3. Use an Existing Azure kubernetes cluster and select the kubernetes cluster that was created during the deploymnet of the resources
  4. Click Next
  5. Specify a Compute Name and click Create

Step 6. Add the Cosmos DB as linked service

In this step you're going to add the Cosmos DB as a linked service in the Synapse Workspace.

  1. Launch the Synapse workspace (via Azure portal > Synapse workspace > Workspace web URL)
  2. Click on Manage - Linked Services - New
  3. Type in the search box "Cosmos" and select the service "Azure Cosmos DB (SQL API)"
  4. Click Continue
  5. Fill in the following data for the linked service
Field Value
Name "patientHubDB"
Connect via integration runtime AutoResolveIntegrationRuntime
Authentication method Connection String
Account selection method Enter Manually
Azure Cosmos DB account URI Copy the Cosmos DB URI from the Cosmos DB you have created
Azure Cosmos DB access key Copy the Cosmos DB Primary Key from the Cosmos DB you have created
Database name "patienthubdb"

Cosmos DB Linked Service Configuration

Step 7. Upload and run Notebooks

  1. Launch the Synapse workspace (via Azure portal > Synapse workspace > Workspace web URL)
  2. Go to Develop, click the +, and click Import to select all Spark notebooks from the repository's /Analytics_Deployment/Synapse-Workspace/Notebooks folder
  3. For each of the notebooks, select Attach to > spark1 in the top dropdown

00_preparedata

  1. Update data_lake_account_name variable to your ADLS in the 00_preparedata.ipynb notebook
  2. Update file_system_name variable to your container in the 00_preparedata.ipynb notebook

01_train_diabetes_readmission_automl

  1. Update data_lake_account_name variable to your ADLS in the 01_train_diabetes_readmission_automl.ipynb notebook
  2. Update file_system_name variable to your container in the 01_train_diabetes_readmission_automl.ipynb notebook
  3. Update subscription_id variable to your Azure SubscriptionID in the 01_train_diabetes_readmission_automl.ipynb notebook
  4. Update resource_group variable to your ResourceGroup in the 01_train_diabetes_readmission_automl.ipynb notebook
  5. Update workspace_name variable to your Azure Machine Learning Workspace in the 01_train_diabetes_readmission_automl.ipynb notebook
  6. Update workspace_region variable to your Azure Machine Learning Workspace Region in the 01_train_diabetes_readmission_automl.ipynb notebook

02_deploy_AKS_diabetes_readmission_model

  1. Update data_lake_account_name variable to your ADLS in the 02_deploy_AKS_diabetes_readmission_model.ipynb notebook
  2. Update file_system_name variable to your container in the 02_deploy_AKS_diabetes_readmission_model.ipynb notebook
  3. Update subscription_id variable to your Azure SubscriptionID in the 02_deploy_AKS_diabetes_readmission_model.ipynb notebook
  4. Update resource_group variable to your ResourceGroup in the 02_deploy_AKS_diabetes_readmission_model.ipynb notebook
  5. Update workspace_name variable to your Azure Machine Learning Workspace in the 02_deploy_AKS_diabetes_readmission_model.ipynb notebook
  6. Update workspace_region variable to your Azure Machine Learning Workspace Region in the 02_deploy_AKS_diabetes_readmission_model.ipynb notebook
  7. Update autoMLRunId variable to the Run ID that was returned in the 01_train_diabetes_readmission_automl notebook in the 02_deploy_AKS_diabetes_readmission_model.ipynb notebook
  8. Update aks_target_name variable to the Compute you have created in step 5 in the 02_deploy_AKS_diabetes_readmission_model.ipynb notebook

When all the variables are modified, publish your all imported notebooks so they are saved in your workspace. Run the following notebooks in order:

Step 8. Deploy API service endpoints on the Azure Kubernetes Service

Before you can confiure the service endpoint, there are some Prerequisites that need to be installed on your device.

  1. Azure CLI - Required for the deployment and configuration of the Azure Resources and Source code
  2. Docker Desktop - Required for Debugging in local or containerizing codes in the Deployment process
  3. PowerShell 7.x - Required to execute the Deployment Script

Open the deployapplications.ps1 file and update the necesary variables.

Field Value
$subscriptionId SubscriptionID where Azure resources will be deployed
$resourcegroupName Resourcegroup Name where Azure resources will be deployed
$containerRegistryName Container Registry Name already deployed in previous deployment by Azure ML
$kubernetesName AKS Name already deployed in previous deployment by Azure ML
$azurecosmosaccountName Azure Cosmos DB Name already deployed in previous deployment
$azurecosmosdbDataBaseName "patientHubDB"
$ttsSubscriptionKey Azure Speech Service Subscription key in previous deployment
$ttsServiceRegion Azure Speech Service Service Region in previous deployment
$mlServiceURL Deployed Realtime inference service URL in Azure ML Studio in previous step
$mlServiceBearerToken Deployed Realtime infrence service BearerToken in Azure ML Studio in previous step

Then run PowerShell as Administrator and launch the script deployapplications.ps1

Once the script is finished, note down the public IP addresses for each services.
You need them in the next step.

Step 9. Run notebook 03_load_predictions

Go back to your Synapse workspace and open the notebook 03_load_predictions Fill in the variables:

  1. Update data_lake_account_name variable to your ADLS in the 03_load_predictions.ipynb notebook
  2. Update file_system_name variable to your container in the 03_load_predictions.ipynb notebook
  3. Update subscription_id variable to your Azure SubscriptionID in the 03_load_predictions.ipynb notebook
  4. Update resource_group variable to your ResourceGroup in the 03_load_predictions.ipynb notebook
  5. Update workspace_name variable to your Azure Machine Learning Workspace in the 03_load_predictions.ipynb notebook
  6. Update workspace_region variable to your Azure Machine Learning Workspace Region in the 03_load_predictions.ipynb notebook

When all the variables are modified, publish your the notebook so it is saved in your workspace, and run it.

Step 10. Deploy and configure the Provider Portal App

  1. Go to https://make.preview.powerapps.com/
  2. In the right upper corner, make sure you select the correct environment where you want to deploy the Power App.
  3. Click on Apps - Import Canvas App
  4. Click upload and select the Frontend_Deployment/PatientHubDoctorPortal.zip Zipfile.
  5. Review the package content. You should see the details as the screenshot below

Review Package Content

  1. Under the Review Package Content, click on the little wrench next to the Application Name Provider Portal, to change the name of the Application. Make sure the name is unique for the environemnt.
  2. Click Import and wait until you see the message All package resources were successfully imported.
  3. Click on Flows. You will notice that all the flows are disabled.

Cloud Flows disabled

  1. You need to turn them on before you can use them. Hover over each of the flows, select the button More Commands and click Turn on.

Turn on Cloud Flows

  1. For each flow, you need to change the HTTP component so that the URI points to your API Services. Edit each flow, open the HTTP component and past the Public IP addresses you noted down in the previous step. Your URI should look similar like the screenshot below.

HTTP

API Service Flow
appointment PatientHub-GetNextAppointments
batchinference PatientHub-InferenceExplanation
patient PatientHub-GetAllPatients
realtimeinference PatientHub-RealtimeInference
tts PatientHub-GetSpeechFile
  1. After the modification, click the "Test" button in the upper right corner to test the flow. If all went well, you should receive "Your flow ran successfully".
  2. Once the flows are modified, you should open the Power App and all should work like a charm.

DISCLAIMER

By accessing this code, you acknowledge that the code is not designed, intended, or made available: (1) as a medical device(s); (2) for the diagnosis of disease or other conditions, or in the cure, mitigation, treatment or prevention of a disease or other conditions; or (3) as a substitute for professional medical advice, diagnosis, treatment, or judgment. Do not use this code to replace, substitute, or provide professional medical advice, diagnosis, treatment, or judgement. You are solely responsible for ensuring the regulatory, legal, and/or contractual compliance of any use of the code, including obtaining any authorizations or consents, and any solution you choose to build that incorporates this code in whole or in part.