STHELAR 40X: Multi-Tissue Histology and Spatial Transcriptomics for Cell Type Prediction

STHELAR (Spatial Transcriptomics and H&E histology for Large-scale Annotation Resource) is a multi-tissue dataset integrating high-resolution spatial transcriptomics data with Hematoxylin & Eosin (H&E) whole slide images. It provides detailed segmentation masks and cell-type annotations for developing models that predict cell types directly from histology. The dataset was authored by FelicieGS and a version containing 587,555 images was last updated on 2026-04-02.

Use Cases

• Training models to predict cell types based on H&E whole slide images.
• Developing segmentation algorithms based on the provided detailed segmentation masks.
• Researching the spatial relationship between gene expression and tissue morphology based on integrated transcriptomics and histology.
• Benchmarking multi-modal learning approaches for medical imaging based on paired image and transcriptomics data.

Strengths

• Integrates two high-resolution data modalities: spatial transcriptomics and H&E histology.
• Provides detailed segmentation masks and cell-type annotations for model training.
• Contains a large volume of 587,555 images in the STHELAR_40x version.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Description metadata is limited; actual data quality requires manual inspection after download.
• The dataset's specific tissue sources, collection protocols, and potential biases are not detailed.

Provenance

Source

huggingface

Collection Method

Integrates high-resolution spatial transcriptomics data with histology whole slide images.

Freshness

Last updated 2026-04-02 14:24:00; freshness should be verified.