Intro to HistomicsTK
This course shows how to create or wrap an image processing or ML workflow so that it can be used by the HistomicsTK / Digital Slide Archive platform, generating editable annotations that can be accessed and used in other tools.
Objectives
- Learn how to start and manage a local instance of the HistomicsTK platform
- Learn how annotations and metadata are created and accessed
- Learn how to wrap an image progressing algorithm to run on one or a batch of histology images
- Learn how ML algorithms can be used
Prerequisites
- Basic Python programming knowledge
- Basic understanding of data processing
- Basic knowledge of ML or some minor knowledge of image processing
- A linux-based computer with docker and docker compose installed on it (optional -- we can provide some web-based computing resources during the course)
Agenda
This course is approximately 8 hours, which can be divided into 2 sessions.
Introduction to HistomicsTK
- History
- Open Source License
- Platform support
- Use-cases
Architecture
- Default docker containers
- Girder for data management
- Slicer CLI execution model
- Task dockers
Managing Data
- Upload / Import files
- Manage permissions
- Annotations (lots of variations)
- Export / Download
- The API (and how to use it from a Jupyter notebook or the command line)
- Running Existing Algorithms
Wrapping a simple image processing task
- A tunable algorithm for foreground / background discrimination for H&E images
- Turn your algorithm into a runnable docker image
- Tile aggregation
- Running the generated docker image on the system
As time permits:
Wrapping a simple ML tensorflow task
- Making an example ML task into a command-line program
- Wrapping it as a docker image
Next Public Course
Registration for training is currently closed. Please check back for future training sessions.
Private Team Training
Our expert instructors can train your team online or at your site. The course syllabus can be customized for your needs.