---
license: cc-by-4.0
---
<div id="top" align="center">

# SurGrID: Controllable Surgical Simulation via Scene Graph to Image Diffusion (IPCAI 2025)

  [![arXiv](https://img.shields.io/badge/arXiv-2502.07945-b31b1b.svg)](https://arxiv.org/abs/2502.07945)
  [![Paper](https://img.shields.io/badge/Paper-Visit-blue)](https://rdcu.be/em4E2)
  [![Hugging Face Spaces](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Spaces-blue)](https://huggingface.co/SsharvienKumar/SurGrID)

</div>


## 💡Key Features
- We show that SGs can encode surgical scenes in a human-readable format. 
- We propose a novel pre-training step that encodes global and local information from (image, mask, SG) triplets. The learned embeddings are employed to condition graph to image diffusion for high-quality and precisely controllable surgical simulation.
- We evaluate our generative approach on scenes from cataract surgeries using quantitative fidelity and diversity measurements, followed by an extensive user study
involving clinical experts


## 🛠 Setup
```bash
git clone https://github.com/MECLabTUDA/SurGrID.git
cd SurGrID
conda create -n surgrid python=3.8.5 pip=20.3.3
conda activate surgrid

pip install torch==2.0.1 torchvision==0.15.2 --index-url https://download.pytorch.org/whl/cu118
pip install -r requirements.txt
```

## 🏁 Model Checkpoints and Dataset
Download the checkpoints of all the necessary models from the provided sources and place them in `[results](./results)`. We also provide the processed CADIS dataset, containing images, segmentation masks and their scene graphs. Update the paths of the dataset in `[configs](./configs)`.
- `Checkpoints`: [VQGANs, GraphEncoders, Diffusion Model](https://huggingface.co/SsharvienKumar/SurGrID/tree/main/checkpoints)
- `Processed Dataset`: [CADIS](https://huggingface.co/SsharvienKumar/SurGrID/tree/main/dataset)


## 💥 Sampling SurGrID
```bash
python script/sampler_diffusion.py --conf configs/eval/eval_combined_emb.yaml
```


## ⏳ Training SurGrID
**Step 1:** Train Separate VQGAN for Image and Segmentation
```bash
python surgrid/taming/main.py --base configs/vqgan/vqgan_image_cadis.yaml -t --gpus 0,
python surgrid/taming/main.py --base configs/vqgan/vqgan_segmentation_cadis.yaml -t --gpus 0,
```

**Step 2:** Train Both Graph Encoder
```bash
python script/trainer_graph.py --mode masked --conf configs/graph/graph_cadis.yaml
python script/trainer_graph.py --mode segclip --conf configs/graph/graph_cadis.yaml
```

**Step 3:** Train Diffusion Model
```bash
python script/trainer_diffusion.py --conf configs/trainer/combined_emb.yaml
```


## 🔄 Training SurGrID on a New Dataset
The files below needs to be adapted:
- [Configs](./configs)
- [SurGrID Dataset](./surgrid/dataset/cadis_dataset.py)
- [VQGAN Dataset](./surgrid/taming/taming/data/cadis.py)
- [CADIS Specifications in Graph Encoder Pre-training](./surgrid/graph/graph_masked_segclip.py)


## 🥼 Clinical Expert Assesment
```bash
python script/demo_surgrid.py --conf configs/trainer/combined_emb.yaml
```
Our demo GUI allows for loading ground-truth graphs along with the ground-truth image. The graph’s nodes can be moved, deleted, or have their class changed. We instruct our participants to load four different ground-truth graphs and sequentially perform the following actions on each. They are requested to score the samples’ realism and coherence with the graph input using a Likert scale of 1 to 7:

- First, participants are instructed to generate a batch of four samples from the groundtruth SG without modifications.
- Second, the participants are requested to spatially move nodes in the canvas and again judge the synthesised samples. 
- Third, participants change the class of one of the instrument nodes and judge the generated images. 
- Lastly, participants are instructed to remove one of the instruments or miscellaneous classes and judge the synthesised image a final time.

<table>
  <thead>
    <tr>
      <th rowspan="2">Clinician</th>
      <th colspan="2">Synthesisation from GT</th>
      <th colspan="2">Spatial Modification</th>
      <th colspan="2">Tool Modification</th>
      <th colspan="2">Tool Removal</th>
    </tr>
    <tr>
      <th>Realism</th>
      <th>Coherence</th>
      <th>Realism</th>
      <th>Coherence</th>
      <th>Realism</th>
      <th>Coherence</th>
      <th>Realism</th>
      <th>Coherence</th>
    </tr>
  </thead>
  <tbody>
    <tr>
      <td>P1</td>
      <td>6.5±0.5</td>
      <td>6.5±1.0</td>
      <td>6.3±0.9</td>
      <td>6.3±0.9</td>
      <td>5.3±1.2</td>
      <td>4.5±1.9</td>
      <td>6.3±0.9</td>
      <td>5.5±2.3</td>
    </tr>
    <tr>
      <td>P2</td>
      <td>5.3±0.9</td>
      <td>5.3±0.5</td>
      <td>4.5±0.5</td>
      <td>4.3±2.0</td>
      <td>5.3±0.9</td>
      <td>5.8±0.9</td>
      <td>5.5±1.2</td>
      <td>5.5±1.9</td>
    </tr>
    <tr>
      <td>P3</td>
      <td>6.3±0.9</td>
      <td>6.3±0.9</td>
      <td>6.5±1.0</td>
      <td>5.5±0.5</td>
      <td>6.0±0.8</td>
      <td>6.8±0.5</td>
      <td>6.3±0.5</td>
      <td>6.5±0.5</td>
    </tr>
  </tbody>
</table>


## 📜 Citations
If you are using SurGrID for your paper, please cite the following paper:
```
@article{frisch2025surgrid,
  title={SurGrID: Controllable Surgical Simulation via Scene Graph to Image Diffusion},
  author={Frisch, Yannik and Sivakumar, Ssharvien Kumar and K{\"o}ksal, {\c{C}}a{\u{g}}han and B{\"o}hm, Elsa and Wagner, Felix and Gericke, Adrian and Ghazaei, Ghazal and Mukhopadhyay, Anirban},
  journal={arXiv preprint arXiv:2502.07945},
  year={2025}
}
```


## ⭐ Acknowledgement
Thanks for the following projects and theoretical works that we have either used or inspired from:
- [VQGAN](https://github.com/CompVis/taming-transformers)
- [Lucidrains' DDPM](https://github.com/lucidrains/denoising-diffusion-pytorch)
- [SGDiff](https://github.com/YangLing0818/SGDiff)
- [Endora's README](https://github.com/CUHK-AIM-Group/Endora)