File size: 18,414 Bytes
ccc243c |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 |
---
license: cc-by-nc-4.0
---
<div align="center">
<h1>Publications</h1>
<h3>Scaling up self-supervised learning for improved surgical foundation models</h3>
[Tim J.M. Jaspers](https://timjaspers0801.github.io/)<sup>1* :email:</sup>, [Ronald L.P.D. de Jong](https://scholar.google.com/citations?user=We226GgAAAAJ&hl=en)<sup>2*</sup>,
[Yiping Li](https://research.tue.nl/nl/persons/yiping-li/publications/)<sup>2</sup>, [Carolus H.J. Kusters](https://chjkusters.github.io/)<sup>1</sup>, Franciscus H.A. Bakker<sup>5</sup>,
Romy C. van Jaarsveld<sup>3</sup>, Gino M. Kuipers<sup>3</sup>, Richard<sup>3</sup>, Jelle P. Ruurda<sup>3</sup>,
Willem M. Brinkman<sup>4</sup>, Josien P.W. Pluim<sup>2</sup>,
Peter H.N. de With<sup>1</sup>, Marcel Breeuwer<sup>2</sup>, [Yasmina Al Khalil](https://scholar.google.com/citations?user=m6co7N0AAAAJ&hl=en)<sup>2</sup>, [Fons van der Sommen](https://scholar.google.com/citations?user=qFiLkCAAAAAJ&hl=en)<sup>1</sup>
<sup>1</sup> Department of Electrical Engineering, Video Coding & Architectures, Eindhoven University of Technology \
<sup>2</sup> Department of Biomedical Engineering, Medical Image Analysis, Eindhoven University of Technology, Eindhoven, The Netherlands \
<sup>3</sup> Department of Surgery, University Medical Center Utrecht, Utrecht, The Netherlands \
<sup>4</sup> Department of Oncological Urology, University Medical Center Utrecht, Utrecht, The Netherlands \
<sup>5</sup> Department of Surgery, Catharina Hospital, Eindhoven, The Netherlands
<sup>*</sup> Both authors attributed equally \
(<sup>:email:</sup>) corresponding author
*arxiv* <br /> ([Article](https://doi.org/10.1016/j.media.2024.103348))
<h3> Exploring the Effect of Dataset Diversity in Self-Supervised Learning for Surgical Computer Vision</h3>
[Tim J.M. Jaspers](https://timjaspers0801.github.io/)<sup>1 :email:</sup>, [Ronald L.P.D. de Jong](https://scholar.google.com/citations?user=We226GgAAAAJ&hl=en)<sup>2</sup>,
[Yasmina Al Khalil](https://scholar.google.com/citations?user=m6co7N0AAAAJ&hl=en)<sup>2</sup>, Tijn Zeelenberg <sup>1</sup>, [Carolus H.J. Kusters](https://chjkusters.github.io/)<sup>1</sup>, Franciscus H.A. Bakker<sup>5</sup>,
[Yiping Li](https://research.tue.nl/nl/persons/yiping-li/publications/)<sup>2</sup>, Romy C. van Jaarsveld<sup>3</sup>, Jelle P. Ruurda<sup>3</sup>, Willem M. Brinkman<sup>4</sup>,
Peter H.N. de With<sup>1</sup>, [Fons van der Sommen](https://scholar.google.com/citations?user=qFiLkCAAAAAJ&hl=en)<sup>1</sup>,
*Second Workshop on Data Engineering in Medical Imaging (DEMI) - Satellite Event MICCAI 2024* <br /> ([Proceeding](https://link.springer.com/chapter/10.1007/978-3-031-73748-0_5))
<sup>1</sup> Department of Electrical Engineering, Video Coding & Architectures, Eindhoven University of Technology \
<sup>2</sup> Department of Biomedical Engineering, Medical Image Analysis, Eindhoven University of Technology, Eindhoven, The Netherlands \
<sup>3</sup> Department of Surgery, University Medical Center Utrecht, Utrecht, The Netherlands \
<sup>4</sup> Department of Oncological Urology, University Medical Center Utrecht, Utrecht, The Netherlands \
<sup>5</sup> Department of Surgery, Catharina Hospital, Eindhoven, The Netherlands
(<sup>:email:</sup>) corresponding author
<h1>Abstract</h1>
Foundation models have revolutionized computer vision by achieving state-of-the-art
performance across diverse tasks through large-scale pretraining on extensive datasets.
However, their application in surgical computer vision has been limited. This study
addresses this gap by introducing SurgeNetXL, a novel surgical foundation model that
sets a new benchmark in surgical computer vision. Trained on the largest reported surgical
dataset to date, comprising over 4.7 million video frames, SurgeNetXL achieves
consistent top-tier performance across six datasets spanning four surgical procedures
and three tasks, including semantic segmentation, phase recognition, and critical view
of safety (CVS) classification. Compared to the best-performing surgical foundation
models, SurgeNetXL shows mean improvements of 2.4, 8.95, and 12.6% for semantic segmentation,
phase recognition, and CVS classification, respectively. Additionally,
SurgeNetXL outperforms the best-performing ImageNet-based variants by 14.4, 4.0,
and 1.6% in the respective tasks. In addition to advancing model performance, this work
provides key insights into scaling pretraining datasets, extending training durations, and
optimizing model architectures specifically for surgical computer vision. These findings
pave the way for improved generalizability and robustness in data-scarce scenarios, offering
a comprehensive framework for future research in this domain.
<h1>Results</h1>
The following figures are from our publications, showcasing the performance of our introduced foundation model
across diverse surgical tasks and procedures. These results demonstrate the model’s state-of-the-art
performance on a variety of downstream tasks, reflecting its versatility and robustness in handling
datasets from multiple surgical procedures.
Figure 1 and Figure 2 illustrate comparative rankings of our model against existing foundation models,
highlighting its superior generalization capabilities across datasets. Figure 3 provides a t-SNE visualization,
showcasing the clear cluster separation per specific dataset achieved by the model’s feature embeddings,
further emphasizing its effectiveness in capturing meaningful representations.
<div style="display: flex; justify-content: space-around; align-items: center; gap: 20px;">
<div style="text-align: center;">
<img src="figures/radar_ranks.png" alt="Fig 2" width="400" height="300">
<p><strong>Fig 1:</strong> Radar chart showing model ranks across datasets.</p>
</div>
<div style="text-align: center;">
<img src="figures/ranking_blob_all.png" alt="Fig 3" width="400" height="300">
<p><strong>Fig 2:</strong> Blob chart representing ranking metrics for models.</p>
</div>
</div>
<div style="text-align: center; margin-top: 20px;">
<img src="figures/TSNE.png" alt="Fig 3" width="600">
<p><strong>Fig 3:</strong> t-SNE visualization of feature embeddings showing cluster separation across datasets.</p>
</div>
<h1>Models</h1>
The models used in this study are based on the [MetaFormer](https://arxiv.org/abs/2210.13452) architecture. The models are trained using a self-supervised learning
approach on the SurgeNetXL dataset and its variations,
introduced this in the following [paper](https://). All model weights can be downloaded from the table below.
| Model | Backbone | Epochs | Teacher Backbone | Full DINO checkpoint |
|-----------------|------------|--------|-------------------------------------------------------------------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------|
| SurgeNetXL | CaFormer | 50 | [Download](https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/SurgeNetXL_checkpoint_epoch0050_teacher.pth?download=true) | [Download](https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/SurgeNetXL_checkpoint0050.pth?download=true) |
| SurgeNetSmall | CaFormer | 50 | [Download](https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/SurgeNetSmall_checkpoint_epoch0050_teacher.pth?download=true) | [Download](https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/SurgeNetSmall_checkpoint0050.pth?download=true) |
| SurgeNetCholec | CaFormer | 50 | [Download](https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/CHOLEC_checkpoint_epoch0050_teacher.pth?download=true) | [Download](https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/CHOLEC_checkpoint0050.pth?download=true) |
| SurgeNetRAMIE | CaFormer | 50 | [Download](https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/RAMIE_checkpoint_epoch0050_teacher.pth?download=true) | [Download](https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/RAMIE_checkpoint0050.pth?download=true) |
| SurgeNetRARP | CaFormer | 50 | [Download](https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/RARP_checkpoint_epoch0050_teacher.pth?download=true) | [Download](https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/RARP_checkpoint0050.pth?download=true) |
| SurgeNetPublic | CaFormer | 50 | [Download](https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/Public_checkpoint_epoch0050_teacher.pth?download=true) | [Download](https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/Public_checkpoint0050.pth?download=true) |
| SurgeNet | CaFormer | 50 | [Download](https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/SurgeNet_checkpoint_epoch0050_teacher.pth?download=true) | [Download](https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/SurgeNet_checkpoint0050.pth?download=true) |
| SurgeNet | ConvNextv2 | 50 | [Download](https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/SurgeNet_ConvNextv2_checkpoint_epoch0050_teacher.pth?download=true) | [Download](https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/SurgeNet_ConvNextv2_checkpoint0050.pth?download=true) |
| SurgeNet | PVTv2 | 50 | [Download](https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/SurgeNet_PVTv2_checkpoint_epoch0050_teacher.pth?download=true) | [Download](https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/SurgeNet_PVTv2_checkpoint0050.pth?download=true) |
<h1>Loading Models</h1>
The weights from the teacher network can be used to initialize either your classification or segmentation model using the following code snippet:
</div>
```python
import torch
from metaformer import caformer_s18, MetaFormerFPN
from convnextv2 import convnextv2_tiny, ConvNextFPN
from pvtv2 import pvt_v2_b2, PVTV2FPN
urls = {
"ImageNet1k": "https://huggingface.co/sail/dl/resolve/main/caformer/caformer_s18.pth",
"SurgeNetXL": "https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/SurgeNetXL_checkpoint_epoch0050_teacher.pth?download=true",
"SurgeNet": "https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/SurgeNet_checkpoint_epoch0050_teacher.pth?download=true",
"SurgeNet-Small": "https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/SurgeNetSmall_checkpoint_epoch0050_teacher.pth?download=true",
"SurgeNet-CHOLEC": "https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/CHOLEC_checkpoint_epoch0050_teacher.pth?download=true",
"SurgeNet-RAMIE": "https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/RAMIE_checkpoint_epoch0050_teacher.pth?download=true",
"SurgeNet-RARP": "https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/RARP_checkpoint_epoch0050_teacher.pth?download=true",
"SurgeNet-Public": "https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/Public_checkpoint0050.pth?download=true",
"SurgeNet-ConvNextv2": "https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/SurgeNet_ConvNextv2_checkpoint_epoch0050_teacher.pth?download=true",
"SurgeNet-PVTv2": "https://huggingface.co/TimJaspersTue/SurgeNetModels/resolve/main/SurgeNet_PVTv2_checkpoint_epoch0050_teacher.pth?download=true",
}
# Metaformer model
classification_model = caformer_s18(num_classes=12, pretrained='SurgeNet', pretrained_weights=urls['SurgeNetXL'])
segmentation_model = MetaFormerFPN(num_classes=12, pretrained='SurgeNet', pretrained_weights=urls['SurgeNetXL'])
# ConvNextv2 model
classification_model = convnextv2_tiny(num_classes=12, pretrained_weights=urls['SurgeNet-ConvNextv2'])
segmentation_model = ConvNextFPN(num_classes=12, pretrained_weights=urls['SurgeNet-ConvNextv2'])
# PVTv2 model
classification_model = pvt_v2_b2(num_classes=12, pretrained_weights=urls['SurgeNet-PVTv2'])
segmentation_model = PVTV2FPN(num_classes=12, pretrained_weights=urls['SurgeNet-PVTv2'])
```
<div align="center">
Note: If your want a different version of SurgeNet weights (e.g. SurgeNet-Small), you can replace the `pretrained_weights` argument with the desired url (leave the `pretrained` argument as it is).
<h1>Surgical Youtube Dataset</h1>
A key contribution of our research is the Surgical YouTube dataset, which enhanced our foundation model's
performance. This curated dataset contains 2,074,234 frames sampled from 23 distinct surgical procedures
and is publicly available at [huggingface datasets.](https://huggingface.co/datasets/TimJaspersTue/SurgeNetYoutube)
This datasets is a large part of our SurgeNetXL dataset, which also includes other opensource datasets.
| Procedure-specific subset | Dataset | Procedure | #videos | #frames | Public |
|---------------------------|----------------------------------------------------------------|-----------|---------|-----------|--------|
| **SurgeNetCholec** | Cholec80 ([Twinnanda et al., 2017b](https://arxiv.org/abs/1602.03012)) | Laparoscopic Cholecystectomy | 76 | 179,164 | Yes |
| | HeiChole ([Maier-Hein et al., 2021](https://www.synapse.org/Synapse:syn25101790/wiki/608802)) | Laparoscopic Cholecystectomy | 30 | 53,427 | Yes |
| | hSDB-Chole ([Yoon et al., 2021](https://arxiv.org/abs/2110.12555)) | Laparoscopic Cholecystectomy | 24 | 18,064 | Yes |
| **SurgeNetRAMIE** | RAMIE-UMCU | RA Esophagectomy | 28 | 377,287 | No |
| **SurgeNetRARP** | ESAD [Bawa et al., 2021](https://arxiv.org/abs/2006.07164) | RA Esophagectomy | 28 | 47,282 | Yes |
| | PSI-AVA [Valderrama et al., 2022](https://arxiv.org/abs/2212.04582) | RA Prostatectomy | 8 | 73,618 | Yes |
| | RARP-AvL | RA Prostatectomy | 8 | 261,516 | No |
| **Others** | DSAD ([Carstens et al., 2023](https://www.nature.com/articles/s41597-022-01719-2)) | RA Rectal Resection/Extirpation | 32 | 14,623 | Yes |
| | GLENDA ([Leibetseder et al., 2020](https://link.springer.com/chapter/10.1007/978-3-030-37734-2_36)) | Gynecologic Laparoscopy | 400 | 25,682 | Yes |
| | LapGyn4 ([Leibetseder et al., 2018](https://dl.acm.org/doi/10.1145/3204949.3208127)) | Gynecologic Laparoscopy | 500 | 59,616 | Yes |
| | MultiBypass140 ([Lavanchy et al., 2024](https://github.com/CAMMA-public/MultiBypass140)) | Laparoscopic Gastric Bypass Surgery | 140 | 749,419 | Yes |
| | hSDB-Gastric ([Yoon et al., 2021](https://arxiv.org/abs/2110.12555)) | RA Gastrectomy | 24 | 35,576 | Yes |
| | SurgToolLoc2022 ([Zia et al., 2023](https://arxiv.org/abs/2305.07152)) | 11 different RA porcine procedures | N/A | 741,516 | Yes |
| | YouTube [ours](https://huggingface.co/datasets/TimJaspersTue/SurgeNetYoutube) | 23 identified procedures | 3,253 | 2,074,234 | Yes |
| SurgeNetXL variations | Dataset | Procedure | #videos | #frames | Public |
|-----------------------|------------------------------------------------------------|---------------------------------------------------------|---------|---------|--------|
| **SurgeNetSmall** | 10% of the above (excluding YouTube) | All of the above (excluding YouTube) | \>1345 | 263,679 | Partly |
| **SurgeNetPublic** | All public datasets (excluding YouTube & private datasets) | All of the above (excluding YouTube & RA Esophagectomy) | \>1238 | 1,997,987 | Yes |
| **SurgeNet** | All of the above (excluding YouTube) | All of the above (excluding YouTube) | \>1345 | 2,636,790 | Partly |
| **SurgeNetXL** | All of the above | All of the above | \>4598 | 4,711,024 | Partly |
<h1>Acknowledgements</h1>
Our implementation of the feature pyramid network is based on the [pytorch segmentation models](https://segmentation-modelspytorch.readthedocs.io/en/latest/) library.
Pretraining on SurgeNet was performed using the code provided with the [DINO](https://github.com/facebookresearch/dino) publication.
We have used the code of Schmidgall et al. (2024) to obtain the youtube videos, this code can be found [here](https://github.com/SamuelSchmidgall/GSViT).
<h1>Citation</h1>
If you find our work useful in your research please consider citing our paper:
</div>
```bibtex
@msc{Jaspers2025,
title={Scaling up self-supervised learning for improved surgical foundation models},
year={2025}
}
```
```bibtex
@inbook{Jaspers2024,
title={Exploring the Effect of Dataset Diversity in Self-supervised Learning for Surgical Computer Vision},
ISBN={9783031737480},
ISSN={1611-3349},
url={http://dx.doi.org/10.1007/978-3-031-73748-0_5},
DOI={10.1007/978-3-031-73748-0_5},
booktitle={Data Engineering in Medical Imaging},
publisher={Springer Nature Switzerland},
author={Jaspers, Tim J. M. and de Jong, Ronald L. P. D. and Al Khalil, Yasmina and Zeelenberg, Tijn and Kusters, Carolus H. J. and Li, Yiping and van Jaarsveld, Romy C. and Bakker, Franciscus H. A. and Ruurda, Jelle P. and Brinkman, Willem M. and De With, Peter H. N. and van der Sommen, Fons},
year={2024},
month=oct, pages={43–53} }
``` |