Orbit: Digital Pathology meets AI (Machine learning & Deep learning) & Big Data with Open-source flavour

Pathology Jul 14, 2019

Orbit is an open source software package built to empower whole slide images (Virtual slides) processing/analysis with powerful AI models for researchers and pathologists.

Orbit is an advanced whole-slide image viewer with advanced image analysis algorithms, built-in machine learning models for tissue quantification, script editor to write, edit models & annotations/RIO tools.

Orbit in action [src: Orbit]


Orbit Supports Whole-slide image formats


Orbit is originally designed to connect and work with image servers like Omero (Whole-slide microscopy image server), but it extended its support to load files from the local machine, including whole-slide image formats, supported formats by Bioformats library, still images format like JPEG, PNG, DICOM format (*.dcm).

Orbit also supports Fluo / multi-channel images, JPEG-XR compression in CZI images & multi-images series with VSI files.

Here is a list of the supported Whole-slide image formats: SVS, NDPI, NDPIS, SCN, AFI, CZI, IMS, VSI, ETS, SLD, TIF, TIFF, TF2, TF8,& BTF.

Orbit: Remote Whole-slide analysis with Virtual-slides server (Omero)


Orbit can open local Whole-slide image in many formats, it is also built to support Omero image server which is an open source microscopy images server that provides a management server app for managing, visualizing, and collaborating on microscopy images with images metadata archiving tools. With Omero support, scientists and researchers will be able to access remote microscopy images and collaborate on Whole-slide image analysis with ease.

Omero is becoming the core Whole-slide image server in the digital pathology industry. It is growing not just because it's an open source project backed by many institutions around the world, but because of its developer-friendly concept, architecture, and tools. It provides the tools for developers & researchers to build digital pathology apps on top of it, and/or integrate it as Orbit does in their software package.

Orbit integrated AI: Machine and Deep learning with Whole-slide analysis

Machine learning and Deep learning are trending AI (Artificial Intelligence)  nowadays, Machine learning is based on the idea that systems (machine) can learn from data, identify patterns and make decisions with minimal human intervention. It is used now in many projects and fields like medical diagnosis, image processing, prediction, object classification, industrial automation, risk assessment, and more.

Deep learning is a subset of machine learning based on the idea that artificial neural networks, algorithms are built to imitate the working of the human brain to process data, create patterns and use them to make a decision. Deep learning is used in many industries and applications right now like Virtual assistants, translations, computer vision (CV), chatbots and automated chat services, facial recognition, marketing industry, medical and pharmaceuticals.

Read: Machine Learning: A gentle introduction
The deep learning algorithm would perform a task repeatedly, each time tweaking it a little to improve the outcome


Orbit has a built-in Sophisticated Image Analysis Algorithms,  machine and deep learning support. The built-in machine learning models can be classification, object segmentation or object classification model. Orbit has a script editor that uses Groovy which provides integration with deep learning model. Orbit's deep learning models can be trained with python scripts.

Q [src: Orbit]
Annotations, & Region of interest (ROI) can be defined by manual annotations or via a trainable exclusion map. Everything can be combined.

Orbit: Big Data & Whole-slide image analysis

Spark or Apache spark is an open source distributed computing and analytics platform for big data and machine learning. As it's built for large-scale data processing, It provides a developer-friendly data processing layout that supports multiple programming languages (Java, Scala, Python, R, and SQL) and play nice with other frameworks, as it provides multiple options to run, & use.

Orbit supports Apache Spark out of the box, which allows the developer to use Spark with IScaleout function. IScaleout can be easily configured but it does not work with standalone mode and requires Omero image server connection.

Orbit in points:

Highlights:

  • Open source true (Libre) software GPLv3
  • Runs on Windows, Linux, macOS
  • Developer-friendly API
  • Sophisticated Image Analysis Algorithms
  • Machine learning
  • Deep Learning ready
  • Big Data-ready with Spark integration
  • OMERO image server support
  • Bio-Formats standard support
  • Scripting support with Groovy
  • Supports high-DPI display (e.g. 4k monitors) Resolution
  • Packed with tutorials, a rich documentation, and beginners handbook.
  • Connectivity
  • Extendability
  • Custom connectors development support: Custom image server

Features:

  • Whole slide image analysis
  • Simple UI with split-pane support
  • Advanced Search
  • History Search
  • Multiple image viewing mode
  • Window/View manager
  • File metadata browser
  • Image manipulation & adjustment
  • Works with files from the local machine or remote images from OMERO image server
  • Powerful viewer tools
  • Script editor: (Groovy)
  • Object manager
  • Tissue quantification using machine learning
  • Object segmentation
  • Object classification
  • Objects training
  • Annotations & RIO (Region of Interest) with annotation tools and options
  • Object detection with Deep learning support
  • Models manager (classification, object segmentation & object classification models)
  • Model: Classes manager
  • Model trainer
  • Deep learning models for detecting arbitrary heterogeneous objects.
  • NATIVE NDPI(S) READER: NDPI and NDPIS files reader with native library (Linux/Windows)
  • Cell Cluster Segmentation
  • Extension manager
  • Cell Profiler (CP) integration for whole slide analysis (Extension)
  • Logging manager
  • A batch tool to apply models to a batch of images
  • Masks Manager
  • Exclusion Models

Extensions:

  • Manual Classification (Extension)
  • Manual Box Count
  • Cell Profiler
  • Nerve Detection
  • TMA Spot Detection
  • Threshold Classification

License:

Orbit is released under GPLv3

Download:

Orbit is available for all known platforms: Windows, Linux, macOS, though it does not provide Linux app packages, Linux users can install it with a setup script.

  • Windows
  • Linux (tar.gz file with installation script)
  • macOS
  • Java version (JAR)

Requirements:

  • Recommended Hardware: Quadcore/Hexacore (e.g. I7/I9), 32 GB RAM (or 64 GB if possible)
  • 64bit OS
  • 4K monitors (high-DPI displays)

Who made it?

Orbit has been developed at Actelion Pharmaceuticals Ltd, now Idorsia Pharmaceuticals Ltd by Manuel Stritt.

Conclusion:

Orbit combines Whole-slide analysis with machine & deep learning packed with the support of big data cluster-computing platform is 100 steps forward for digital pathology.

It's built for researchers with an interest in building and training AI models, which saves a huge deal of time and effort to integrate and build such a tool. Let's imagine the price of such software if it is released as a proprietary commercial product instead of an open source public license project.

We recommend it for researchers, pathologists, data scientists and software engineers who are into big data, machine learning & deep learning to give it a try.


Whole-slide Image Analysis

Open source Free Whole-slide Image Viewers and Analysis Software

PathologyAnalyticsBig DataWhole-slide ImageVirtual SlideOpen sourceData scienceData Analysis


Hamza Mu

A physician with programming skills, Linux user since late 1990s, Open source supporter . Doing coding with Python, NodeJS (Meteor, VueJS, Express, D3, PhantomJS), SmallTalk & R language.