histomicsui 1.4.2.dev14

Organize, visualize, and analyze histology images.

Organize, visualize, and analyze histology images.

HistomicsUI organizes and manages whole slide image (WSI) files using Girder. It has a dedicated interface to select WSI, add annotations manually, and to run analysis and algorithms on all or parts of images.

Girder provides authentication, access control, and diverse storage options, including using local file systems and Amazon S3. WSI images are read and displayed via the large_image module. Algorithms are containerized using Docker and are run using the slicer_cli_web Girder plugin. These can be run on multiple worker machines via Girder Worker and celery.

A set of common algorithms are provided by HistomicsTK.

Installation

Linux

In linux with Python 3.6 or newer:

Prerequisites:

• MongoDB must be installed and running.

• An appropriate version of Python must be installed.

pip install histomicsui[analysis] --find-links https://girder.github.io/large_image_wheels
girder build
girder serve

To use Girder Worker:

pip install girder_slicer_cli_web[worker]
GW_DIRECT_PATHS=true girder-worker -l info -Ofair --prefetch-multiplier=1

The first time you start HistomicsUI, you’ll also need to configure Girder with at least one user and one assetstore (see the Girder documentation). Additionally, it is recommended that you install the HistomicsTK algorithms. This can be done going to the Admin Console, Plugins, Slicer CLI Web settings. Set a default task upload folder, then import the dsarchive/histomicstk:latest docker image.

Reference Deployment

The standard deployment of HistomicsUI is the Digital Slide Archive. The associated repository has tools for readily installing via Docker, VirtualBox, or shell scripts on Ubuntu.

Development

The most convenient way to develop on HistomicsUI is to use the devops scripts from the Digital Slide Archive.

If you are making changes to the HistomicsUI frontend, you can make Girder watch the source code and perform hot reloads on changes using the --watch-plugin argument to girder build. See the Girder docs for more information.

Annotations and Metadata from Jobs

This handles ingesting annotations and metadata that are uploaded and associating them with existing large image items in the Girder database. These annotations and metadata re commonly generated through jobs, such as HistomicTK tasks, but can also be added manually.

If a file is uploaded to the Girder system that includes a reference record, and that reference record contains an identifier field and at least one of a fileId and an itemId field, specific identifiers can be used to ingest the results. If a userId is specified in the reference record, permissions for adding the annotation or metadata are associated with that user.

Metadata

Identifiers ending in ItemMetadata are loaded and then set as metadata on the associated item that contains the specified file. Conceptually, this is the same as calling the PUT item/{id}/metadata endpoint.

Annotations

Identifiers ending in AnnotationFile are loaded as annotations, associated with the item that contains the specified file. Conceptually, this is the same as uploaded the file via the annotation endpoints for the item associated with the specified fileId or itemId.

If the annotation file contains any annotations with elements that contain girderId values, the girderId values can be identifier values from files that were uploaded with a reference record that contains a matching uuid field. The uuid field is required for this, but is treated as an arbitrary string.

Funding

This work was funded in part by the NIH grant U24-CA194362-01.