CardiacSeg

Requirements

• miniconda
• python 3.9

Install

• install packages.

./setup.sh

• setup data and model dir.

python setup_dir.py

• set config to config.toml, for download data and model.

# dataset google dirve file id
[dataset]
chgh='<google_drive_id>'
mmwhs='<google_drive_id>'

# model google dirve file id
[model.chgh]
unet3d='<google_drive_id>'
attention_unet='<google_drive_id>'
cotr='<google_drive_id>'
unetr='<google_drive_id>'
swinunetr='<google_drive_id>'
unetcnx_a1='<google_drive_id>'

[model.mmwhs]
unet3d='<google_drive_id>'
attention_unet='<google_drive_id>'
cotr='<google_drive_id>'
unetr='<google_drive_id>'
swinunetr='<google_drive_id>'
unetcnx_a1='<google_drive_id>'

• download data and model.

python download_data.py

Training/Test

Open Notebook

• open training notebook from CardiacSeg/exps/exp_chgh.ipynb.

Setup Config

workspace

• setup absolute path of workspace.

workspace = '<workspace>/CardiacSeg'

model name

• setup model name.
• The model name used in this study is unetcnx_a1.
• If you want to replace it with other research methods, you can change it to a different model name, such as swinunetr, unetr, cotr, attention_unet and unet.

model_name = 'unetcnx_a1'

Run

• run all cells, and the final results of the program will display validation scores and inference scores.

Infer

Open Notebook

• open training notebook from CardiacSeg/exps/infer_chgh.ipynb.

Setup Config

workspace

• setup absolute path of workspace.

workspace = '<workspace>/CardiacSeg'

model name

• setup model name.
• The model name used in this study is unetcnx_a1.
• If you want to replace it with other research methods, you can change it to a different model name, such as swinunetr, unetr, cotr, attention_unet and unet.

model_name = 'unetcnx_a1'

exp name

• setup model name.
• If you want to replace it with other research methods, such as swinunetr, unetr, cotr, attention_unet and unet. you can change it to a different exp name 't_4'.

exp_name = 't_5'

pid

• setup pid (patient id).

pid = 'pid_1000'

Run

• after the inference is completed, the program will output the inference result and display the path of the output result (last line).

Download

• download inference result

Show

• display the inference results using 3D Slicer.

CardiacLab

• integrate MONAI Label with 3D Slicer.
• https://github.com/conflick0/CardiacLab