Spatial Transcriptomics
Cross-source consensus on Spatial Transcriptomics from 1 sources and 5 claims.
1 sources · 5 claims
How it works
Risks & contraindications
Comparisons
Evidence quality
Where it comes from
Highlighted claims
- MIST derives its molecular prototypes from HEST-1k, a dataset of 1,276 spatially-resolved transcriptomic profiles paired with H&E WSIs across 26 organ types and approximately 940,000 Visium spots. — Bridging the Modality Bottleneck in Pathology MIL through Virtual Molecular Staining
- Spatial transcriptomics bridges H&E morphology and molecular state by co-registering spot-level gene expression with the underlying tissue image. — Bridging the Modality Bottleneck in Pathology MIL through Virtual Molecular Staining
- Prior ST-augmented computational pathology methods incorporate molecular signal by fine-tuning or re-pretraining the pathology foundation model with vision-omics contrastive objectives. — Bridging the Modality Bottleneck in Pathology MIL through Virtual Molecular Staining
- Fine-tuning-based ST approaches tightly couple representations to a specific FM checkpoint, requiring costly retraining whenever a new foundation model is released. — Bridging the Modality Bottleneck in Pathology MIL through Virtual Molecular Staining
- MIST prototype affinity maps closely match ground-truth transcriptomic prototype maps on held-out HEST slides, indicating that morphological signals genuinely correlate with molecular programs. — Bridging the Modality Bottleneck in Pathology MIL through Virtual Molecular Staining