Medical Imaging Segmentation Toolkit

About

The Medical Imaging Segmentation Toolkit (MIST) is a simple, scalable, and end-to-end 3D medical imaging segmentation framework. MIST allows researchers to seamlessly train, evaluate, and deploy state-of-the-art deep learning models for 3D medical imaging segmentation.

Please cite the following papers if you use this code for your work:

A. Celaya et al., "PocketNet: A Smaller Neural Network For Medical Image Analysis," in IEEE Transactions on Medical Imaging, doi: 10.1109/TMI.2022.3224873.

A. Celaya et al., "FMG-Net and W-Net: Multigrid Inspired Deep Learning Architectures For Medical Imaging Segmentation," in Proceedings of LatinX in AI (LXAI) Research Workshop @ NeurIPS 2023, doi: 10.52591/lxai202312104

A. Celaya et al. "MIST: A Simple and Scalable End-To-End 3D Medical Imaging Segmentation Framework," arXiv preprint arXiv:2407.21343

Documentation

Please see our Read the Docs page here.

What's New

• October 2024 - MIST takes 3rd place in BraTS 2024 adult glioma challenge @ MICCAI 2024!
• August 2024 - Added clDice as an available loss function.
• April 2024 - The Read the Docs page is up!
• March 2024 - Simplify and decouple postprocessing from main MIST pipeline.
• March 2024 - Support for using transfer learning with pretrained MIST models is now available.
• March 2024 - Boundary-based loss functions are now available.
• Feb. 2024 - MIST is now available as PyPI package and as a Docker image on DockerHub.
• Feb. 2024 - Major improvements to the analysis, preprocessing, and postprocessing pipelines, and new network architectures like UNETR added.
• Feb. 2024 - We have moved the TensorFlow version of MIST to mist-tf.