Model summary of iStar. The histology image is hierarchically divided into tiles, which are then converted into hierarchical histology image features. These features, combined with the spot-level gene expression data, are then utilized to predict super-resolution gene expression.