Data Sheet 1_SpatialFinder: a human-in-the-loop vision-language framework for prioritizing

A 2.7 MB dataset released by Jonathan Xu on April 15, 2026, containing the SpatialFinder framework. It is a human-in-the-loop vision-language model designed to predict gene expression heterogeneity and rank high-value regions of interest from H&E tissue slides. The framework was evaluated across four Visium HD tissue types.

Use Cases

• Prioritizing high-value regions of interest for spatial transcriptomics assays based on H&E slide analysis.
• Benchmarking vision-language models for predicting gene expression heterogeneity in tissue samples.
• Training human-in-the-loop AI systems for clinical decision support in pathology.

Strengths

• Framework achieved Spearman's ρ up to 0.89 and Overlap@10% up to 78.8% in evaluations.
• Demonstrated a 24.9 percentage-point gain over the strongest vision-language model baseline.
• Released under a permissive CC-BY-4.0 license.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect bias inherent to the specific four Visium HD tissue types used for evaluation.

Provenance

Source

figshare

Collection Method

Likely contains the code, model, and evaluation data for the SpatialFinder framework as described.

Time Range

null

Freshness

Last updated 2026-04-15 05:22:59; freshness should be verified.

Geography

null