File size: 2,482 Bytes
e570dbe
 
 
 
 
 
 
 
 
 
 
 
 
 
3d9c7e7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license: mit
task_categories:
- image-feature-extraction
- image-segmentation
- image-classification
language:
- en
tags:
- biology
pretty_name: STimage-1K4M
size_categories:
- 100B<n<1T
---
# STimage-1K4M Dataset

Welcome to the STimage-1K4M Dataset repository. This dataset is designed to foster research in the field of spatial transcriptomics, combining high-resolution histopathology images with detailed gene expression data. 

![teaser](aux/f1.png "teaser")


## Dataset Description

STimage-1K4M consists of 1,149 spatial transcriptomics slides, totaling over 4 million spots with paired gene expression data. This dataset includes:

- Images.
- Gene expression profiles matched with high-resolution histopathology images.
- Spatial coordinates for each spot.

## Data structure
The data structure is organized as follows:

```bash
β”œβ”€β”€ annotation              # Pathologist annotation
β”œβ”€β”€ meta                    # Test files (alternatively `spec` or `tests`)
β”‚   β”œβ”€β”€ bib.txt             # the bibtex for all studies with pmid included in the dataset
β”‚   β”œβ”€β”€ meta_all_gene.csv   # The meta information
β”œβ”€β”€ ST                      # Include all data for tech: Spatial Transcriptomics
β”‚   β”œβ”€β”€ coord               # Include the spot coordinates & spot radius of each slide
β”‚   β”œβ”€β”€ gene_exp            # Include the gene expression of each slide
β”‚   └── image               # Include the image each slide
β”œβ”€β”€ Visium                  # Include all data for tech: Visium, same structure as ST
β”œβ”€β”€ VisiumHD                # Include all data for tech: VisiumHD, same structure as ST
```
## Repository structure

The code for data processing and reproducing evaluation result in the paper are in [Document](https://jiawenchenn.github.io/STimage-1K4M/docs/01-make-meta).

## Acknowledgement
The fine-tuning and evaluation codes borrows heavily from [CLIP](https://github.com/openai/CLIP/issues/83) and [PLIP](https://github.com/PathologyFoundation/plip/). 

## Citation

```
@misc{chen2024stimage1k4m,
      title={STimage-1K4M: A histopathology image-gene expression dataset for spatial transcriptomics}, 
      author={Jiawen Chen and Muqing Zhou and Wenrong Wu and Jinwei Zhang and Yun Li and Didong Li},
      year={2024},
      eprint={2406.06393},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}
```

## License

All code is licensed under the MIT License - see the LICENSE.md file for details.