GitHub tutorials

Stay organized with collections Save and categorize content based on your preferences.

This page contains tutorials and codelabs hosted on GitHub that demonstrate how to create applications with the Cloud Healthcare API.

Getting the tutorials

The latest Cloud Healthcare API tutorials are available in the GoogleCloudPlatform/healthcare GitHub repository.

Understanding the tutorials

The provided tutorials give concrete examples of how the Cloud Healthcare API is used. In the next section, you'll be introduced to each tutorial and given information on what the tutorial does and what parts of the Cloud Healthcare API it features.

Several of the tutorials use public datasets that are provided by the Cloud Healthcare API for use with your applications.

DICOM tutorials

The following tutorials demonstrate end-to-end machine learning solutions for DICOM medical imaging using the Cloud Healthcare API and other Google Cloud products.

Breast density classification

Two training and inference on breast density classification model tutorials are provided to train, deploy, and run inference on a breast density classification model. Breast density is a possible factor in the risk for breast cancer. Both tutorials use the TCIA CBIS-DDSM (Curated Breasted Imaging Subset of DDSM) dataset. Both tutorials show how to store, retrieve, and transcode medical images that use the DICOM format.

The tutorials both run on Vertex AI Workbench user-managed notebooks. One tutorial uses AutoML Vision for machine learning, and the other uses AI Platform.

FHIR tutorials

The following tutorials demonstrate end-to-end solutions for FHIR workflows, including machine learning and data discovery.

FHIR immunizations tutorial

The FHIR immunizations tutorial shows how to:

  • Create a Cloud Healthcare API web application that implements a FHIR workflow.
  • Use TensorFlow to make predictions about whether patients will develop a disease.
  • Export the FHIR data to BigQuery and explore the data.

The tutorial involves working with a dynamically generated dataset that contains patient information.

What's next