Automatic generation of pathological benchmark dataset from hyperspectral images of double stained tissues

•A strategy is presented for generating annotated pathological benchmark datasets from hyperspectral images of double-stained tissues.•Avoided mismatches between tissues and annotations on separate stained sections by H&E-IHC composite staining.•Spatial-Spectral based Hyperspectral GAN is designed to eliminate staining differences and establish a dataset.•Established the spectral prior into GBDT to automatically generate annotation and improve segmentation accuracy.