Skip to main content

Welcome to the Frontier Fusion Developer Portal

Discover developer resources​

The remainder of this chapter gives an overview of what the Frontier Fusion platform is and what role the developer portal plays in this context.

Getting Started will explain the basics of building a prototype for Frontier Fusion.

In addition there are several tutorials which go into details of creating and onboarding different types of prototypes. Start with Python based prototype and walk your way through the remaining tutorials.

info

For SHS internal early adopter users: Please use this teams channel to interact with the developer portal team if you need help or have any feedback or bug report! We are happy to help and appreciate all your feedback! (The teams channel should be accessibel to anybody. If you are experiencing issues, please contact sven.bauer@siemens-healthineers.com)

Immediate access to clinical users​

The "Frontier Fusion" research environment enables researchers to easily onboard, deploy and update prototypes. It is targeted at the radiological domain and hence focuses on the processing of DICOM data.

One of the main benefits of using the "Frontier Fusion" environment is that it provides immediate access to clinical users, which already have a syngo.via or a teamplay environment (with the respective Frontier license). In other words: These clinical users can start working with the newly onboarded prototype within minutes. More specifically, users with a syngo.via environment (or other OpenApps compatible environment) can directly download and install the prototype through the "Marketplace". teamplay users can utilize the prototype by simply enabling it in their "AI-Rad Companion Research" configuration.

This gives researchers a fast-lane to clinical evaluation of algorithms and other prototypes with the necessary data routing infrastructure being taken care of.

High level overview​

One of the fundamental ideas of the platform is that prototypes will be developed once and then deployed into multiple different target environments (currently two are supported, more planned). With this approach we try to maximize the reach of a prototype into the clinical research user base. Ultimately, feedback from research users is elemental in improving the prototype and eventually turning it into a product.

In particular, we are currently supporting the following deployment options:

  • Cloud: The prototype is deployed into the "AI-Rad Companion Research" environment. It is similar to the "AI Rad Companion" product offering and runs in the Azure cloud. The connection to the users environment will be made through the teamplay platform. Further details below.
    • This environment is ideal for customers who prefer minimal effort and cost on their side and have no objections to sending data to the cloud for processing.
  • Edge: The "AI Rad Companion Research" environment also supports an "edge" deployment option where data stays on the users premises but management is done centrally form the cloud. Furthermore, diagnostic information is sent back to the cloud (logs) for central error analysis. This variant is combining the advantages of an on-premises and a cloud deployment. However, it has higher demands for the locally running hardware.
    • This environment is ideal for customers who prefer not to send their data to the cloud for processing but still want to have the system centrally managed by SHS.
  • syngo.via OpenApps: The prototype is deployed as a so-called OpenApps on a syngo.via system (or other OpenApps capable system).
    • This environment is ideal for customers who already have a syngo.via system and want to utilize it for additional processing.

The following image shows a high level architecture overview of the AIRC Research environment:

AIRC Research Overview

TODO: Refine this image and make it specific for this project and integrate open apps.

Please refer to the Getting Started to build and onboard your first prototype!

Develop once, run multiple​

TOOD: Explain how we wrap the prototypes to deploy it to open apps and AIRC research

Role of the developer portal​

TODO

Advanced and future topics​

Topics in this chapter are work in progress and only mentioned to complete the picture and provide a future outlook. They are not available yet in the context of the Frontier Fusion developer portal.

Data sharing​

The Frontier Fusion platform allows users of prototypes to share data with the developer. This is closing the data loop and enables model re-training. Data sharing will be enabled in a later version of the platform.

Using Azure Machine Learning Studio for retraining your models​

Frontier Fusion integrates seamlessly with Azure Machine Learning Studio. This will allow for a professional development environment for your machine learning tasks while still maintaining the connection with the developer portal for the operational tasks.