HistomicsML2

HistomicsML2 is a software platform for fast and interactive development of deep learning classifiers from whole-slide imaging datasets. Scientists can use the browser-based interface of HistomicsML2 to train and validate classifiers for patterns like tumor infiltrating lymphocytes to support tissue-based studies. An approach called active learning guides users to label the most valuable training instances, producing more accurate classifiers with less time and effort.

Getting started

HistomicsML2 is provided as a collection of Docker images that simplify system deployment and dataset creation. Follow the quick start guide to download and run containers with pre-loaded sample data and see the section on training classifiers.

For a more detailed overview of HistomicsML2 see system overivew.

To create HistomicsML2 dataset from your own images and to create a system deployment see creating datasets for HistomicsML2 and importing HistomicsML2 datasets.

To format data from your own image segmentation and feature extraction algorithms for use with HistomicsML2 see data formats.

Table of contents

• Quick start guide
• Training classifiers with HistomicsML2
• Inference and exporting results
• System overview
• Creating datasets for HistomicsML2
• Importing datasets
• Data formats

Resources

• Project source: HistomicsML2 (https://github.com/CancerDataScience/HistomicsML2).

• Related projects: HistomicsTK (https://github.com/DigitalSlideArchive/HistomicsTK).