2026-10-10T12:59:15Z https://www.scidb.cn/oai

10.57760/sciencedb.44563 2026-08-03T17:31:00Z

2026-08-03 NSCLC-HEMIF: A large-scale paired H&E and multiplex immunofluorescence image dataset for virtual staining of the non-small cell lung cancer microenvironment doi:10.57760/sciencedb.44563 en NSCLC-HEMIF: A paired H&E and multiplex immunofluorescence image datasetNSCLC-HEMIF is a curated paired histopathology image dataset developed for virtual staining and cross-modal computational pathology research in the non-small cell lung cancer microenvironment. The dataset was generated from adjacent serial tissue sections stained with hematoxylin and eosin (H&E) and multiplex immunofluorescence (mIF). The mIF panel includes the nuclear counterstain DAPI and five immune markers: CD4, CD20, CD38, CD66B and FOXP3.The released dataset contains 10,000 coordinate-matched patch sets, each with a spatial resolution of 270 × 270 pixels. Each patch set comprises eight JPG images: one H&E patch, one mIF composite patch, one DAPI patch and five marker-specific patches corresponding to CD4, CD20, CD38, CD66B and FOXP3. The files are organized into eight modality-specific folders, and identically named files across these folders represent the same tissue region from the same WSI-level chip and spatial coordinate. An accompanying metadata spreadsheet links each retained patch to its patient- and slide-level information.The data-generation workflow included whole-slide image registration, spectral unmixing, marker-wise morphological denoising, coordinate-matched patch extraction, automated effective-signal filtering using a 1.0% threshold and expert pathology-guided curation. Technical validation assessed patient-level patch distribution, H&E–mIF structural correspondence using normalized mutual information, hematoxylin–DAPI nuclear-pattern consistency and marker-wise fluorescence signal-to-noise ratios.NSCLC-HEMIF is intended to support the development and benchmarking of virtual staining, image-to-image translation and spatial immune-marker prediction methods. The paired H&E, mIF composite, DAPI and marker-specific channels may also support studies of image registration, nuclear-pattern analysis and multimodal representation learning. Non-small cell lung cancer; Virtual staining; Hematoxylin and eosin; Multiplex immunofluorescence; Computational pathology Xu Lu 罗国栋 Qingyang Wang Shaopeng Liu Xuan Wu Wenhua Liang EMBARGO https://creativecommons.org/licenses/by-nc-nd/4.0/ dataset Science Data Bank