Spatial transcriptomics inferred from pathology whole-slide images links tumor heterogeneity to survival in breast and lung cancer
Problem
State-of-the-art diagnostics lack automated methods to characterize and quantify tumor heterogeneity in a spatial context from pathology whole-slide images.
Why it matters
Tumor heterogeneity is an important determinant of cancer prognosis and treatment matching.
What the AI did
The authors trained deep learning to infer spatial transcriptomics from pathology H&E whole-slide images, addressing heterogeneity characterization without requiring many bulk training samples or complex pipelines.
Results
not stated in source
Impact
Title claims image-derived tumor heterogeneity links to survival in breast and lung cancer; abstract truncated before details.
Record
| Venue | Scientific Reports |
|---|---|
| Published | 2020 |
| Work type | article |
| Authors | Alona Levy-Jurgenson, Xavier Tekpli, Vessela Nedelcheva Kristensen, Zohar Yakhini |
| Citations | 155 (OpenAlex, at import time) |
| References | 77 |
| Model or method | deep learning (specific architecture not stated) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41598-020-75708-z |
| Open access | gold |
| Keywords | spatial transcriptomics, whole-slide images, tumor heterogeneity, deep learning, cancer survival |
Importance rating
significant — thousands to millions affected
Tumor heterogeneity is an important determinant of cancer prognosis and treatment matching; models reach 0.95 AUC on breast and lung cancer cohorts.