NA-MIC Project Weeks
Back to Projects List
Slicer On Demand
Key Investigators
- Steve Pieper (Isomics, Inc. Cambridge MA, USA)
- Curt Lisle (Knowledgevis, Maitland, Florida, USA)
- Andrey Fedorov (BWH, Boston, MA, USA)
- Theodore Aptekarev (Independent, Moscow, Russia)
Project Description
The goal is to allow people to quickly transition from viewing images to doing more complex tasks such as segmentation or registration.
Objective
- A quick and easy way to get a functioning Slicer environment
- Ability to browse data, e.g. in IDC, and load same data in cloud-hosted Slicer
- Pass login credentials from web to Slicer to allow load/save of confidential data
Approach and Plan
- Build on existing SlicerMachines GCP boot images
- Prototype using the IDC sandbox that already supports login
- Use GCP JavaScript API to launch and monitor jobs
Progress and Next Steps
- Steve implemented one-click creation of a VM using the Google API to launch a GPU-enabled VM, ready to use within about 90 seconds.
- Theodore created workflow icons to provide visual feedback during the launch process.
- We met and discussed methods for encrypting traffic to the “pop up” Slicer-in-the-cloud using Google Cloud infrastructure options.
- A use case was identified where an IDC cohort manifest could be passed to the Slicer VM and the cohort could be automatically loaded for the user.
- We had a discussion with Kitware regarding the composition of the Slicer Docker containers: It would be nice to consolidate dockerfile of general use into https://github.com/Slicer/SlicerDocker
- Next Steps:
- Evaluate tradeoffs between simplicity of interface and exposing options
- Test robustness, add more feedback about things like how much money you are spending
- Configure the VM instance with tools and ML models
- Improve the desktop/window managment setup to be more modern
Illustrations
| Click on the image below to see the video |
|---|
 |
 |