moose-v.2.2.35

[Bug] Included the right opencv-python version

https://github.com/QIMP-Team/MOOSE/issues/67

Moose v2.2.33 Release Notes

Date: 3rd october 2023

---

New Features

• Custom nnU-Net Trainer: Specialized trainer for optimized vertebrae segmentation.
• Severe Data Augmentation (DA5): Enhanced data augmentation strategy for improved generalization.
• No Mirroring: Focused on capturing clinically relevant features by removing data mirroring.

---

Notes

• Backward-compatible but upgrading is strongly recommended.
• Performance metrics comparing new and old models to be released soon.

moose v2.2.32 Release Notes

Release Date

2023-09-26

Overview

We are excited to announce the release of moose v2.2.32! This update brings new features and improvements that make medical image analysis for preclinical arena easier and more efficient.

What's New

Features

Preclinical CT Legs Model

• We've included a new model called preclin_ct_legs, designed to segment mouse preclinical CT images for the left and right leg.

Improvements

tqdm Progress Bar in nnunetv2

• The tqdm progress bar from nnunetv2 is now hidden. We rerouted it to stderr to make the user interface cleaner.

Installation

To install the latest version, run the following command:

pip install moosez==2.2.32

or update using:

pip install --upgrade moosez

Acknowledgments

We would like to thank our contributors and users for their continued support and feedback.

MOOSE 2.0 is here!

🎉 Announcing MOOSE 2.0: Leaner. Meaner. Stronger 🎉

Exciting times are ahead! We're thrilled to unveil MOOSE 2.0, taking 3D medical image segmentation to unprecedented heights! 🚀

🌟 Features at a Glance:

• Leaner: Optimized for efficiency, MOOSE 2.0 doesn't demand extensive resources. It's compatible with various OS and even works without high-end GPUs (a bit slower though).

• Meaner: A remarkable speed upgrade – we're talking about a version that's 5x faster than version 1! Designed for both clinical and preclinical (coming soon) settings, this is a segmentation powerhouse. ⚡

• Stronger: The strength of MOOSE 2.0 is undeniable, backed by Data-centric AI principles and a staggering 2.5k datasets.

• Versatility: Whether you prefer command-line tools for batch processing or using it as a library for your Python projects, MOOSE 2.0 offers unmatched flexibility. 😎

📌 Ready to Dive In?
Before you start, ensure you meet the requirements:

• OS Compatibility: Windows, Mac, or Linux.
• Memory: At least 32GB RAM.
• GPU: For enhanced speed, an NVIDIA GPU is recommended.
• Python: Version 3.9 or above.

🔧 Quick Installation:

• For Linux and MacOS:

  python3 -m venv moose-env
  source moose-env/bin/activate
  pip install moosez

• For Windows:

  python -m venv moose-env
  .\moose-env\Scripts\activate
  pip install moosez

✨ How to Use:

• As a Command-Line Tool:

  moosez -d <path_to_image_dir> -m <model_name>

• As a Library in Python Projects:

  from moosez import moose
  moose(model_name, input_dir, output_dir, accelerator)

📂 Adherence to the specified directory structure and naming conventions is crucial for the best results with MOOSE 2.0.

🎁 Contribute to MooseZ:
Join the MooseZ community! Add your custom nnUNetv2 models to MooseZ and enjoy the speed and efficiency it offers.

🔍 The 'Z' in our Python Packages:
Our signature 'Z' is a testament to our innovative spirit at QIMP. It signifies our quest for the unknown, always pushing the boundaries in medical imaging.

Dive into the complete README for a detailed exploration. Here's to redefining the future of medical image segmentation! Join us in this exhilarating journey with MOOSE 2.0. 🚀🔬

Happy segmenting! 💡🎊

moose-v0.1.4

What's Changed

• Feature: Adding checks for environment variables by @LalithShiyam in #43
• Bug: nnUNet broke suddenly due to version issues, now MOOSE installation file will always build the latest version of nnUNet from the git repo (MIC-DKFZ/nnUNet#1132)! Please re-install MOOSE, if MOOSE doesn't work due to this bug.

Full Changelog: moose-v0.1.3...moose-v0.1.4

moosev0.1.3

What's Changed

• Created CODE_OF_CONDUCT.md by @LalithShiyam in #32
• Updated README.md by @LalithShiyam in #35
• Created a docker image for MOOSEv0.1.0 by @LalithShiyam in #37

Full Changelog: moose-v0.1.2...moose-v0.1.3

moose-v0.1.2-rc

Full Changelog: moose-v0.1.1-rc...moose-v0.1.2

MOOSE-v0.1.1-rc

What's Changed

• BUG: Fixed moose_uninstaller to remove env variables. by @LalithShiyam in #28

Full Changelog: moose-v0.1.0-rc...moose-v0.1.1-rc

MOOSE v0.1.0-rc

What's Changed

• The source code has been made modular to ensure maintainability.
• MOOSE now generates log files for each run, which makes it easier to debug.
• The output messages are much cleaner and organised, with clean progress bars.
• FSL dependency is completely removed. We use nibabel now.
• MOOSE now creates a stats folder which contains the following metrics in a '.csv' file:

• SUV (mean, max, std, max, min) values, if PET images are provided

• HU units (mean, max, std, max, min)

• Volume metrics from CT

• MOOSE now has a binary classifier (fastai-based) which figures out if a given PET volume has a brain in the field-of-view, works most of the times.
• Automated affine alignment between PET/CT, if both images are present. Just to ensure spatial alignment.

New Contributors

• @LalithShiyam made their first contribution in #4
• @Keyn34 made their first contribution in #11

Full Changelog: https://github.com/QIMP-Team/MOOSE-v0.1.0/commits/moose-v0.1.0-rc

** To-do:

• Docker image for the current version