MOOSE

Add a comment

Multi-organ objective segmentation

MOOSE (Multi-organ objective segmentation) is a data-centric AI solution designed to streamline systemic Total-Body research by generating multilabel organ segmentations with high precision. Built on the robust nn-UNet framework, MOOSE can identify 120 unique tissue classes from whole-body 18F-FDG PET/CT images, accepting inputs as either combined PET/CT images or CT-only images.

The data-centric AI principles lie at the core of MOOSE’s innovation. By maintaining a fixed state-of-the-art architecture (nnUNet), the training dataset is augmented to optimize segmentation performance. MOOSE leverages the concept of similarity space to systematically monitor and integrate data that enhances performance. MOOSE is designed to harness the full potential of total-body PET datasets.

Features (MOOSE 2.0 version)

  • requires less than 32GB of RAM, maintains compatibility across operating systems and provides flexibility to work with or without NVIDIA GPUs,
  • attains a speed clocking of 5x faster than its predecessor, MOOSE 1.0, serving up a range of segmentation models designed for both clinical and preclinical settings without much time compromise,
  • employs a data-centric AI approach utilizing 1.5K whole-body PET/CT datasets (~40x times more data than previous version) to improve precision, outcomes and reliability,
  • offers the choices of either a powerful command-line interface for batch processing, or of a library package for individual processing in Python projects,
  • accommodates an array of modalities including PET, CT, and MRI

MOOSE is developed by the QIMP lab members of the Medical University of Vienna and is part of the ENHANCE.PET initiative.

GitHub Page

Installation Guide

User’s Guide

Developer’s Guide

Blog Site

References

  1. Shiyam Sundar, L. K., Yu, J., Muzik, O., Kulterer, O., Fueger, B. J., Kifjak, D., Nakuz, T., Shin, H. M., Sima, A. K., Kitzmantl, D., Badawi, R. D., Nardo, L., Cherry, S. R., Spencer, B. A., Hacker, M., & Beyer, T. (2022). Fully-automated, semantic segmentation of whole-body 18F-FDG PET/CT images based on data-centric artificial intelligence. Journal of Nuclear Medicine. https://doi.org/10.2967/jnumed.122.264063
  2. Isensee, F., Jaeger, P.F., Kohl, S.A.A. et al. nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation. Nat Methods 18, 203–211 (2021). https://doi.org/10.1038/s41592-020-01008-z

Leave a comment