Jointly leveraging spatial transcriptomics and deep learning models for pathology image annotation improves cell type identification over either approach alone.