Pathology Benchmark · MahmoodLab
hest-bench
A pathology benchmark evaluating spatial-transcriptomics and pathology models on aligned histology and expression tasks.
What hest-bench is
hest-bench is a pathology benchmarking dataset developed by MahmoodLab. It is designed as a companion benchmark to HEST-1k to evaluate machine learning systems in biological and medical domains.
The benchmark focuses on evaluating spatial-transcriptomics and pathology foundation models on tasks that pair histological imaging with gene expression data.
How it was made
It supports processing via data science and machine learning data loaders such as Polars, Dask, and Hugging Face Datasets.
Key facts
- Category
- Pathology Benchmark
- Fields
- Medicine, biology, benchmark
- Data modalities
- Image, text
- File format
- CSV
- Related resource
- HEST-1k
What you can use it for
- Pathology foundation model evaluation
- Benchmarking the performance of pathology foundation models on tasks involving tissue histology.
- Spatial-transcriptomics model evaluation
- Assessing spatial-transcriptomics models on tasks requiring joint interpretation of morphology and expression.
- Histology-expression alignment benchmarking
- Testing multimodal computational systems on aligned histology and transcriptomic expression tasks.
Explore the data
No licence declared
The authors have not published a licence, which means no permission to redistribute the data has been granted to anyone. The rows are not served here; read them on the original dataset page.How the columns are distributed
Over 187 rows of default/train
expr_path
string_text- min
- 15
- median
- 16
- mean
- 16.40
- max
- 18
patches_path
string_text- min
- 15
- median
- 16
- mean
- 16.40
- max
- 18
sample_id
string_text- min
- 4
- median
- 5
- mean
- 5.40
- max
- 7
Strengths and limits
Good at
- Targets evaluation across aligned histology and spatial-transcriptomics tasks.
- Serves as a standardised companion evaluation suite for HEST-1k.
- Compatible with data analysis tools including Polars, Dask, and Datasets.
Watch out for
- Requires evaluation methodologies compatible with paired image and text-based tabular data.
Questions
Can I look inside hest-bench without downloading it?
Not here: it declares no licence, so no permission to redistribute its rows has been granted. Its description and statistics are still shown.
What licence is hest-bench under?
Its authors have not declared a licence, which means no permission to reuse or redistribute it has been granted.
What is the primary objective of hest-bench?
It provides a benchmark to test spatial-transcriptomics and pathology foundation models on aligned histology and gene expression tasks.
What is the relationship between hest-bench and HEST-1k?
hest-bench serves as a companion benchmark to the HEST-1k dataset.
What data modalities are present in hest-bench?
The benchmark includes image and text modalities distributed in CSV format.
Written from hest-bench's own dataset card. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.