{"id":997,"date":"2021-10-06T08:38:58","date_gmt":"2021-10-06T08:38:58","guid":{"rendered":"https:\/\/salarydistribution.com\/machine-learning\/2021\/10\/06\/build-a-system-for-catching-adverse-events-in-real-time-using-amazon-sagemaker-and-amazon-quicksight\/"},"modified":"2021-10-06T08:38:58","modified_gmt":"2021-10-06T08:38:58","slug":"build-a-system-for-catching-adverse-events-in-real-time-using-amazon-sagemaker-and-amazon-quicksight","status":"publish","type":"post","link":"https:\/\/salarydistribution.com\/machine-learning\/2021\/10\/06\/build-a-system-for-catching-adverse-events-in-real-time-using-amazon-sagemaker-and-amazon-quicksight\/","title":{"rendered":"Build a system for catching adverse events in real-time using Amazon SageMaker and Amazon QuickSight"},"content":{"rendered":"
\n

Social media platforms provide a channel of communication for consumers to talk about various products, including the medications they take. For pharmaceutical companies, monitoring and effectively tracking product performance provides customer feedback about the product, which is vital to maintaining and improving patient safety. However, when an unexpected medical occurrence resulting from a pharmaceutical product administration occurs, it\u2019s classified as an adverse event (AE). This includes medication errors, adverse drug reactions, allergic reactions, and overdoses. AEs can happen anywhere: in hospitals, long-term care settings, and outpatient settings.<\/p>\n

The objective of this post is to provide an example that showcases how to use Amazon SageMaker<\/a> and pre-trained transformer models to detect AEs mentioned on social media. The model is fine-tuned on domain-specific data to perform a text classification task. We also use Amazon QuickSight<\/a> to create a monitoring dashboard. Importantly, this post requires a Twitter developer account to obtain tweets. For the purposes of this demonstration, we only use publicly available tweets. While privacy and data governance is not explicitly discussed in this post, users should consider these processes and helpful resources can be found in AWS Marketplace<\/a> and through featured AWS Partner Solutions<\/a> for data governance. Following this demonstration, we deleted all the data used.<\/p>\n

This post is meant to support overarching pharmacovigilance activities for the life sciences and pharmaceutical customers globally, though the reference architecture can be implemented for any customer. The model is trained on identifying adverse events and can be applicable to biotech, healthcare, and life sciences domains.<\/p>\n

Solution overview<\/h2>\n

The following architecture diagram illustrates the workflow of the solution.
\"\"<\/a>
The workflow includes the following steps:<\/p>\n

    \n
  1. Train a classification model using SageMaker training and deploy the trained model to an endpoint using SageMaker real-time inference.<\/li>\n
  2. Create an AWS Cloud9<\/a> stream listener.<\/li>\n
  3. Store live tweets in Amazon DynamoDB<\/a>.<\/li>\n
  4. Use DynamoDB Streams to trigger AWS Lambda<\/a> to invoke the SageMaker endpoint for AE classification and call Amazon Comprehend Medical<\/a> to detect symptoms and provide ICD-10<\/a> descriptions.<\/li>\n
  5. Save tweets with their AE classification and symptoms in Amazon Simple Storage Service<\/a> (Amazon S3) and use the AWS Glue<\/a> Data Catalog to create table views.<\/li>\n
  6. Analyze data from Amazon S3 using Amazon Athena<\/a>.<\/li>\n
  7. Create a QuickSight dashboard to monitor tweets and their AE status.<\/li>\n<\/ol>\n

    Set up the environment and implement the solution<\/h2>\n

    We have created a template for the adverse event detection app using the AWS Cloud Development Kit<\/a> (AWS CDK), an open-source software development framework to define your cloud application resources. Complete the following steps to run the solution end to end:<\/p>\n

      \n
    1. Clone the GitHub repo<\/a> either to your local machine that is configured with your AWS account and has the AWS Command Line Interface<\/a> (AWS CLI) installed, or in an AWS Cloud9 environment within your AWS account:\n
      \n
      $ git clone https:\/\/github.com\/aws-samples\/aws-cdk-adverse-event-detection-app<\/code><\/pre>\n<\/p><\/div>\n<\/li>\n<\/ol>\n
      After you clone the code repo, you can start the deployment process.<\/p>\n
        \n
      1. Navigate to the project directory and create a virtual environment within this project that is stored under the .venv<\/code> directory. To manually create a virtual environment on macOS or Linux, use the following code:\n          <\/li>\n
      2. After the virtual environment is created, activate the virtual environment:\n
        \n
        $ source .venv\/bin\/activate<\/code><\/pre>\n<\/p><\/div>\n<\/li>\n
      3. After the virtual environment is activated, install the required dependencies:\n
        \n
        $ pip install -r requirements.txt<\/code><\/pre>\n<\/p><\/div>\n<\/li>\n
      4. At this point, you can now synthesize the AWS CloudFormation<\/a> template:\n
        \n
        $ cdk synth\n$ cdk deploy --all<\/code><\/pre>\n<\/p><\/div>\n<\/li>\n<\/ol>\n
        cdk synth<\/code> generates the CloudFormation template in JSON format as well as other necessary asset files for spinning up the resources. These files are stored in the cdk.out<\/code> directory. The cdk deploy<\/code> command then deploys the stack into your AWS account. You deploy two stacks: one is an S3 bucket stack and the other is the core adverse event app stack. The core app stack needs to be deployed after the Amazon S3 stack is successfully deployed. If you encounter any issues during the deployment process, refer to Troubleshooting common AWS CDK issues<\/a>.<\/p>\n
        After the AWS CDK is successfully deployed, you need to train and deploy a model. On the Notebooks <\/strong>page of the SageMaker console, you should find a notebook instance named AdverseEventDetectionModeling<\/code>. When you run the entire notebook (AE_model_train_deploy.ipynb<\/a>), a SageMaker training job is launched and the model is deployed to a SageMaker endpoint. The model training data in this tutorial is based on the Adverse Drug Reaction<\/a> dataset from Hugging Face<\/a> but can be replaced with any other dataset.<\/p>\n
        Train and deploy a transformer model for adverse event classification<\/h2>\n
        We fine-tune transformer models within the Hugging Face library<\/a> for adverse event (AE) classifications. The training job is built using the SageMaker PyTorch estimator. For model deployment, we use PyTorch Model Server. In this section, we walk through the major steps for model training and deployment.<\/p>\n
        Data preparation<\/h3>\n
        We use the Adverse Drug Reaction Data (ade_corpus_v2<\/a>) within the Hugging Face dataset as the training and validation data. The required data structure for our model training and inference has two columns:<\/p>\n