HESpotEx: a dual-stream deep learning framework for spot-level gene expression prediction from histological images - PubMed
a day ago
- #histological image analysis
- #deep learning
- #spatial transcriptomics
- HESpotEx is a dual-stream deep learning framework that predicts spot-level gene expression from histological images.
- It uses a graph attention autoencoder, an image encoder, and a graph convolution network decoder to predict expressions of up to 5,457 genes from whole-slide images.
- The framework shows superior performance, robustness across cancer and non-cancer samples, and identifies diagnosis-associated patches in in-house datasets.
- HESpotEx demonstrates better cross-sectional consistency in high-resolution spatial transcriptomics datasets, highlighting its potential to link histological patterns with molecular characteristics.