in this study, we propose a Hierarchical ST variational autoencoder (HiSTaR) to extract multi-level latent features of spots. HiSTaR tends to perform well in identifying spatial domains across multiple datasets from diverse platforms, consistently showing superior results compared to existing methods. HiSTaR downstream analysis dataset. including 10x Genomics Visium, Slide-seqV2, Stereo-seq, and STARmap