Pixel2Cancer

Cellular Automata in Computed Tomography

This repository provides the code and checkpoints for our novel tumor synthesis approach, Pixel2Cancer, which can simulate tumor development within organs with realistic texture, shape, and interactions with other tissues.

Paper

From Pixel to Cancer: Cellular Automata in Computed Tomography
Yuxiang Lai^1,2, Xiaoxi Chen³, Angtian Wang¹, Alan L. Yuille¹, and Zongwei Zhou^1,*
1 Johns Hopkins University
² Southeast University,
³ University of Illinois Urbana-Champaign
International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI 2024; Early Accept)
paper | code

We have summarized publications related to tumor synthesis in Awesome Synthetic Tumors .

Model

Organ	Tumor	Model	Pre-trained?	Download
liver	real	unet	no	link
liver	real	swin_unetrv2_base	no	link
liver	synt	unet	no	link
liver	synt	swin_unetrv2_base	no	link
pancreas	real	unet	no	link
pancreas	real	swin_unetrv2_base	no	link
pancreas	synt	unet	no	link
pancreas	synt	swin_unetrv2_base	no	link
kidney	real	unet	no	link
kidney	real	swin_unetrv2_base	no	link
kidney	synt	unet	no	link
kidney	synt	swin_unetrv2_base	no	link

You can download other materials from these links:

All other checkpoints: link

Data: Liver (link), Kidney (link), Pancreas (link)

0. Installation

git clone https://github.com/MrGiovanni/Pixel2Cancer.git
cd Pixel2Cancer/
# download pre-trained models
wget https://github.com/Project-MONAI/MONAI-extra-test-data/releases/download/0.8.1/model_swinvit.pt

See installation instructions to create an environment and obtain requirements.

1. Train segmentation models using synthetic tumors

datapath=/mnt/zzhou82/PublicAbdominalData/

# UNET (no.pretrain)
CUDA_VISIBLE_DEVICES=0,1,2,3 python -W ignore main.py --optim_lr=4e-4 --batch_size=2 --lrschedule=warmup_cosine --optim_name=adamw --model_name=unet --val_every=200 --max_epochs=2000 --save_checkpoint --workers=0 --noamp --distributed --dist-url=tcp://127.0.0.1:12235 --cache_num=200 --val_overlap=0.5 --syn --logdir="runs/synt.no_pretrain.unet" --train_dir $datapath --val_dir $datapath --json_dir datafolds/healthy.json

# Swin-UNETR-Base (pretrain)
CUDA_VISIBLE_DEVICES=0,1,2,3 python -W ignore main.py --optim_lr=4e-4 --batch_size=2 --lrschedule=warmup_cosine --optim_name=adamw --model_name=swin_unetrv2 --swin_type=base --val_every=200 --max_epochs=2000 --save_checkpoint --workers=0 --noamp --distributed --dist-url=tcp://127.0.0.1:12231 --cache_num=200 --val_overlap=0.5 --syn --logdir="runs/synt.pretrain.swin_unetrv2_base" --train_dir $datapath --val_dir $datapath --json_dir datafolds/healthy.json --use_pretrained

# Swin-UNETR-Base (no.pretrain)
CUDA_VISIBLE_DEVICES=0,1,2,3 python -W ignore main.py --optim_lr=4e-4 --batch_size=2 --lrschedule=warmup_cosine --optim_name=adamw --model_name=swin_unetrv2 --swin_type=base --val_every=200 --max_epochs=2000 --save_checkpoint --workers=0 --noamp --distributed --dist-url=tcp://127.0.0.1:12231 --cache_num=200 --val_overlap=0.5 --syn --logdir="runs/synt.no_pretrain.swin_unetrv2_base" --train_dir $datapath --val_dir $datapath --json_dir datafolds/healthy.json

# Swin-UNETR-Small (no.pretrain)
CUDA_VISIBLE_DEVICES=0,1,2,3 python -W ignore main.py --optim_lr=4e-4 --batch_size=2 --lrschedule=warmup_cosine --optim_name=adamw --model_name=swin_unetrv2 --swin_type=small --val_every=200 --max_epochs=2000 --save_checkpoint --workers=0 --noamp --distributed --dist-url=tcp://127.0.0.1:12233 --cache_num=200 --val_overlap=0.5 --syn --logdir="runs/synt.no_pretrain.swin_unetrv2_small" --train_dir $datapath --val_dir $datapath --json_dir datafolds/healthy.json

# Swin-UNETR-Tiny (no.pretrain)
CUDA_VISIBLE_DEVICES=0,1,2,3 python -W ignore main.py --optim_lr=4e-4 --batch_size=2 --lrschedule=warmup_cosine --optim_name=adamw --model_name=swin_unetrv2 --swin_type=tiny --val_every=200 --max_epochs=2000 --save_checkpoint --workers=0 --noamp --distributed --dist-url=tcp://127.0.0.1:12234 --cache_num=200 --val_overlap=0.5 --syn --logdir="runs/synt.no_pretrain.swin_unetrv2_tiny" --train_dir $datapath --val_dir $datapath --json_dir datafolds/healthy.json

2. Train segmentation models using real tumors (for comparison)

datapath=/mnt/zzhou82/PublicAbdominalData/

# UNET (no.pretrain)
CUDA_VISIBLE_DEVICES=0,1,2,3 python -W ignore -W ignore main.py --optim_lr=4e-4 --batch_size=2 --lrschedule=warmup_cosine --optim_name=adamw --model_name=unet --val_every=200 --val_overlap=0.5 --max_epochs=2000 --save_checkpoint --workers=2 --noamp --distributed --dist-url=tcp://127.0.0.1:12235 --cache_num=200 --logdir="runs/real.no_pretrain.unet" --train_dir $datapath --val_dir $datapath --json_dir datafolds/lits.json

# Swin-UNETR-Base (pretrain)
CUDA_VISIBLE_DEVICES=0,1,2,3 python -W ignore -W ignore main.py --optim_lr=4e-4 --batch_size=2 --lrschedule=warmup_cosine --optim_name=adamw --model_name=swin_unetrv2 --swin_type=base --val_every=200 --val_overlap=0.5 --max_epochs=2000 --save_checkpoint --workers=2 --noamp --distributed --dist-url=tcp://127.0.0.1:12231 --cache_num=200 --logdir="runs/real.pretrain.swin_unetrv2_base" --train_dir $datapath --val_dir $datapath --json_dir datafolds/lits.json --use_pretrained

# Swin-UNETR-Base (no.pretrain)
CUDA_VISIBLE_DEVICES=0,1,2,3 python -W ignore -W ignore main.py --optim_lr=4e-4 --batch_size=2 --lrschedule=warmup_cosine --optim_name=adamw --model_name=swin_unetrv2 --swin_type=base --val_every=200 --val_overlap=0.5 --max_epochs=2000 --save_checkpoint --workers=2 --noamp --distributed --dist-url=tcp://127.0.0.1:12232 --cache_num=200 --logdir="runs/real.no_pretrain.swin_unetrv2_base" --train_dir $datapath --val_dir $datapath --json_dir datafolds/lits.json

# Swin-UNETR-Small (no.pretrain)
CUDA_VISIBLE_DEVICES=0,1,2,3 python -W ignore -W ignore main.py --optim_lr=4e-4 --batch_size=2 --lrschedule=warmup_cosine --optim_name=adamw --model_name=swin_unetrv2 --swin_type=small --val_every=200 --val_overlap=0.5 --max_epochs=2000 --save_checkpoint --workers=2 --noamp --distributed --dist-url=tcp://127.0.0.1:12233 --cache_num=200 --logdir="runs/real.no_pretrain.swin_unetrv2_small" --train_dir $datapath --val_dir $datapath --json_dir datafolds/lits.json

# Swin-UNETR-Tiny (no.pretrain)
CUDA_VISIBLE_DEVICES=0,1,2,3 python -W ignore -W ignore main.py --optim_lr=4e-4 --batch_size=2 --lrschedule=warmup_cosine --optim_name=adamw --model_name=swin_unetrv2 --swin_type=tiny --val_every=200 --val_overlap=0.5 --max_epochs=2000 --save_checkpoint --workers=2 --noamp --distributed --dist-url=tcp://127.0.0.1:12234 --cache_num=200 --logdir="runs/real.no_pretrain.swin_unetrv2_tiny" --train_dir $datapath --val_dir $datapath --json_dir datafolds/lits.json

3. Evaluation

AI model trained by synthetic tumors

datapath=/mnt/zzhou82/PublicAbdominalData/

# UNET (no.pretrain)
CUDA_VISIBLE_DEVICES=0 python -W ignore validation.py --model=unet --val_overlap=0.75 --val_dir $datapath --json_dir datafolds/lits.json --log_dir runs/synt.no_pretrain.unet --save_dir out
# Swin-UNETR-Base (pretrain)
CUDA_VISIBLE_DEVICES=0 python -W ignore validation.py --model=swin_unetrv2 --swin_type=base --val_overlap=0.75 --val_dir $datapath --json_dir datafolds/lits.json --log_dir runs/synt.pretrain.swin_unetrv2_base --save_dir out
# Swin-UNETR-Base (no.pretrain)
CUDA_VISIBLE_DEVICES=0 python -W ignore validation.py --model=swin_unetrv2 --swin_type=base --val_overlap=0.75 --val_dir $datapath --json_dir datafolds/lits.json --log_dir runs/synt.no_pretrain.swin_unetrv2_base --save_dir out
# Swin-UNETR-Small (no.pretrain)
CUDA_VISIBLE_DEVICES=0 python -W ignore validation.py --model=swin_unetrv2 --swin_type=small --val_overlap=0.75 --val_dir $datapath --json_dir datafolds/lits.json --log_dir runs/synt.no_pretrain.swin_unetrv2_small --save_dir out
# Swin-UNETR-Tiny (no.pretrain)
CUDA_VISIBLE_DEVICES=0 python -W ignore validation.py --model=swin_unetrv2 --swin_type=tiny --val_overlap=0.75 --val_dir $datapath --json_dir datafolds/lits.json --log_dir runs/synt.no_pretrain.swin_unetrv2_tiny --save_dir out

AI model trained by real tumors

datapath=/mnt/zzhou82/PublicAbdominalData/

# UNET (no.pretrain)
CUDA_VISIBLE_DEVICES=0 python -W ignore validation.py --model=unet --val_overlap=0.75 --val_dir $datapath --json_dir datafolds/lits.json --log_dir runs/real.no_pretrain.unet --save_dir out
# Swin-UNETR-Base (pretrain)
CUDA_VISIBLE_DEVICES=0 python -W ignore validation.py --model=swin_unetrv2 --swin_type=base --val_overlap=0.75 --val_dir $datapath --json_dir datafolds/lits.json --log_dir runs/real.pretrain.swin_unetrv2_base --save_dir out
# Swin-UNETR-Base (no.pretrain)
CUDA_VISIBLE_DEVICES=0 python -W ignore validation.py --model=swin_unetrv2 --swin_type=base --val_overlap=0.75 --val_dir $datapath --json_dir datafolds/lits.json --log_dir runs/real.no_pretrain.swin_unetrv2_base --save_dir out
# Swin-UNETR-Small (no.pretrain)
CUDA_VISIBLE_DEVICES=0 python -W ignore validation.py --model=swin_unetrv2 --swin_type=small --val_overlap=0.75 --val_dir $datapath --json_dir datafolds/lits.json --log_dir runs/real.no_pretrain.swin_unetrv2_small --save_dir out
# Swin-UNETR-Tiny (no.pretrain)
CUDA_VISIBLE_DEVICES=0 python -W ignore validation.py --model=swin_unetrv2 --swin_type=tiny --val_overlap=0.75 --val_dir $datapath --json_dir datafolds/lits.json --log_dir runs/real.no_pretrain.swin_unetrv2_tiny --save_dir out

Data Setting

Train on 9k data of AbdominalAtlas1.1:

The release of AbdomenAtlas 1.0 can be found at https://huggingface.co/datasets/AbdomenAtlas/AbdomenAtlas_1.0_Mini

# AbdominalAtlas1.1 training data list
# Liver 
--json_dir /datafolds/Bodymap/Atlas9k_liver/Atlas9k_liver_fold0.json
--json_dir /datafolds/Bodymap/Atlas9k_liver/Atlas9k_liver_fold1.json
--json_dir /datafolds/Bodymap/Atlas9k_liver/Atlas9k_liver_fold2.json
--json_dir /datafolds/Bodymap/Atlas9k_liver/Atlas9k_liver_fold3.json
--json_dir /datafolds/Bodymap/Atlas9k_liver/Atlas9k_liver_fold4.json
#Pancreas
--json_dir /datafolds/Bodymap/Atlas9k_liver/Atlas9k_pancreas_fold0.json
--json_dir /datafolds/Bodymap/Atlas9k_liver/Atlas9k_pancreas_fold1.json
--json_dir /datafolds/Bodymap/Atlas9k_liver/Atlas9k_pancreas_fold2.json
--json_dir /datafolds/Bodymap/Atlas9k_liver/Atlas9k_pancreas_fold3.json
--json_dir /datafolds/Bodymap/Atlas9k_liver/Atlas9k_pancreas_fold4.json
#Kidney
--json_dir /datafolds/Bodymap/Atlas9k_liver/Atlas9k_kidney_fold0.json
--json_dir /datafolds/Bodymap/Atlas9k_liver/Atlas9k_kidney_fold1.json
--json_dir /datafolds/Bodymap/Atlas9k_liver/Atlas9k_kidney_fold2.json
--json_dir /datafolds/Bodymap/Atlas9k_liver/Atlas9k_kidney_fold3.json
--json_dir /datafolds/Bodymap/Atlas9k_liver/Atlas9k_kidney_fold4.json

Train on public data & Intern experiments:

# Public training data list
# Liver
--json_dir /datafolds/5_fold/liver/liver_tumor_0.json
--json_dir /datafolds/5_fold/liver/liver_tumor_1.json
--json_dir /datafolds/5_fold/liver/liver_tumor_2.json
--json_dir /datafolds/5_fold/liver/liver_tumor_3.json
--json_dir /datafolds/5_fold/liver/liver_tumor_4.json
# Pancreas
--json_dir /datafolds/5_fold/pancreas/pancreas_tumor_0.json
--json_dir /datafolds/5_fold/pancreas/pancreas_tumor_1.json
--json_dir /datafolds/5_fold/pancreas/pancreas_tumor_2.json
--json_dir /datafolds/5_fold/pancreas/pancreas_tumor_3.json
--json_dir /datafolds/5_fold/pancreas/pancreas_tumor_4.json
# Kidney
--json_dir /datafolds/5_fold/kidney/kidney_tumor_0.json
--json_dir /datafolds/5_fold/kidney/kidney_tumor_1.json
--json_dir /datafolds/5_fold/kidney/kidney_tumor_2.json
--json_dir /datafolds/5_fold/kidney/kidney_tumor_3.json
--json_dir /datafolds/5_fold/kidney/kidney_tumor_4.json

Acknowledgement

This work was supported by the Lustgarten Foundation for Pancreatic Cancer Research and the McGovern Foundation. The segmentation backbone is based on Swin UNETR; we appreciate the effort of the MONAI Team to provide and maintain open-source code to the community.