Auto Segmentation models
Clone this wiki locally
CERR includes a library of deep Learning-based auto-segmentation models. The models are distributed as Singularity containers or packaged conda environments to ensure reproducibility across operating systems. Models for auto-segmenting the following sites/modalities/organs are available. Submit request to get access to Singularity Container or Conda Environment for Windows, Mac or Linux OS.
|Imaging Site||Modality||Organ/s||DL Model/ Framework||Reference||Apps in Container|
|LUNG||CT||Heart, Heart Structures, Pericardium, Atria, Ventricles||DeepLab, Pytorch||Haq et al, Physics and Imaging in Radiation Oncology, Vol 14, pp 61-66, 2020 https://doi.org/10.1016/j.phro.2020.05.009||CT_Atria_DeepLab, CT_Heart_DeepLab, CT_HeartStructure_DeepLab, CT_Pericardium_DeepLab, CT_Ventricles_DeepLab|
|LUNG||CT||Nodules||incremental MRRN, Keras, Tensorflow||Jiang et al, IEEE Transactions on Medical Imaging, 38(1): 134 – 144 https://doi.org/10.1109/TMI.2018.2857800||CT_Lung_incrMRRN|
|Prostate||MRI||Bladder, Prostate and Seminal Vesicles (CTV), Penile Bulb, Rectum, Urethra and Rectal Spacer||DeepLabV3+, Tensorflow-GPU||Elguindi et al, Physics and Imaging in Radiation Oncology, Vol.12, pp. 80-86, 2019, https://doi.org/10.1016/j.phro.2019.11.006||MR_Prostate_DeepLab|
|Head & Neck||CT||Parotids (left, right), Submandibulars (left, right), Mandible, Brain Stem, Spinal cord, Oral cavity and larynx||Self Attention, Pytorch||Jiang et al, https://arxiv.org/abs/1909.05054||CT_HeadAndNeck_SelfAttention|
|Head & Neck||CT||Masseters (left, right), medial pterygoids (left, right), pharyngeal constrictor muscle and larynx||DeepLabV3+, Pytorch||Iyer et al, https://doi.org/10.1088/1361-6560/ac4000||CT_ChewingStructures_DeepLabV3, CT_Larynx_DeepLabV3, CT_PharyngealConstrictor_DeepLabV3
Running CERR segmentation models in local Anaconda environment (YouTube)
Detailed instructions to run the above algorithms
Boilerplate for deploying models in CERR
To provide guidance for incorporating homegrown deep learning segmentation models as CERR pipelines, we have created boilerplate code to assist with training and container creation.
Tips on resolving errors while using Conda Environments
There might be conflicts with system dependencies while running Conda Environments. Here is advice on resolving some of the known issues:
|ImportError: DLL load failed: The operating system cannot run %1.||remove C:\Windows\System32\libiomp5md.dll|
The codebase for implementations of models uses the GNU-GPL copyleft license (https://www.gnu.org/licenses/lgpl-3.0.en.html) to allow open-source distribution with additional restrictions. The license retains the ability to propagate any changes to the codebase back to the open-source community along with the following restrictions (i) No Clinical Use, (ii) No Commercial Use, and (iii) Dual Licensing which reserve the right to diverge and/or modify and/or expand the model implementations library to have a closed source/proprietary version along with the open source version in future. We would like to highlight that the library of implementations presented in this work is not approved by the U.S. Food and Drug Administration and should not be used to make clinical decisions for treating patients. The library merely provides implementations of the developed models, whereas the creators of models retain the copyright to their work.
Aditya P. Apte, Aditi Iyer, Maria Thor, Rutu Pandya, Rabia Haq, Jue Jiang, Eve LoCastro, Amita Shukla-Dave, Nishanth Sasankan, Ying Xiao, Yu-Chi Hu, Sharif Elguindi, Harini Veeraraghavan, Jung Hun Oh, Andrew Jackson, Joseph O. Deasy, Library of deep-learning image segmentation and outcomes model-implementations, Physica Medica, Volume 73, 2020, Pages 190-196, ISSN 1120-1797, https://doi.org/10.1016/j.ejmp.2020.04.011.