Overview Segmentation of Total Metabolic Tumor Volume (TMTV) holds significant value in facilitating quantitative imaging biomarkers for lymphoma management. We addressed the challenging task of automated tumor delineation in lymphoma from PET/CT scans. Our model, TMTV-Net, is specifically designed to automate TMTV segmentation from PET/CT scans. TMTV-Net demonstrates robust performance and adaptability in TMTV segmentation across diverse multi-site external datasets, encompassing various lymphoma subtypes. Furthermore, we have containerized our model and made it available in this repository for additional multi-site evaluations and generalizability analyses.

We welcome and encourage you to share your results with us. (Email: [email protected])

Figure 1: TMTV-Net, was utilized at the University of Wisconsin on their dataset for the segmentation of TMTV. (a) Hodgkin Case, DSC=0.83, TMTV relative error=0.18, (b) DLBCL case, DSC=0.66, TMTV relative error=0.10, (c) Hodgkins case, DSC=0.76, (d) DLBCL, DSC=0.67.

#📁 Required folder structure for training

For training you need to convert the nifti files to HDF5:

cd <root>/auto_pet/src/auto_pet/projects/segmentation/preprocessing/
python create_dataset.py

The preprocessed data will be stored here: <root>/dataset/preprocessed/

This repository includes a Docker setup for running inference using TMTV-Net. The Docker image encapsulates the necessary environment, dependencies, and the trained model for seamless and reproducible inference on new data.

• src: Python files serving as libraries and helpers.
• models: Folder containing the trained model weights (.model files).
• dockerfile: Configuration file for building the Docker image.
• main.py: Entry point for running inference.
• requirements.txt: List of Python dependencies required for the Docker image.

Before running inference, build the Docker image:

``bash docker build -t tmtv-net-inference .

Run the Docker container for inference:

docker run -it tmtv-net-inference python main.py

📦 Git LFS (Large File Storage) Due to the large size of the model files, we use Git LFS (Large File Storage) to efficiently handle and version these files. Make sure you have Git LFS installed to fetch the model weights properly.

Installing Git LFS:

sudo apt-get install git-lfs

brew install git-lfs

Clone Repository with LFS: git lfs install git clone https://github.com/qurit-frizi/TMTV-Net.git

Feel free to explore and adapt the provided commands based on your specific folder structure or naming conventions. This Docker setup ensures a consistent and reproducible environment for running TMTV-Net inference.

TMTV-Net is shared for research-use only. COMMERCIAL USE IS PROHIBITED for the time being. For further information please email [email protected]

Please cite the following paper if you use TMTV-Net for your research: Yousefirizi, F. et. al. TMTV-Net: Fully Automated Total Metabolic Tumor Volume Segmenta-tion in Lymphoma PET/CT Images – a Multi-Center Generalizability Analysis, European Journal of Nuclear Medicine and Molecular Imaging. 2024 Feb 8. doi: 10.1007/s00259-024-06616-x.

Torch Research Workflows

We welcome and encourage you to share your results with us.