2026-10-15T01:02:41Z https://www.scidb.cn/oai

10.57760/sciencedb.37165 2026-06-11T10:43:01Z

2026-06-11 Ganzhou Municipal Hospital Gland dataset (GZMH_GLAND) doi:10.57760/sciencedb.37165 en Quantitative assessment of glandular formation is a critical parameter in breast cancer diagnosis. While deep learning methods can assist pathologists, datasets specifically targeting breast cancer pathological images remain scarce. Existing datasets often originate from academic challenges, exhibiting limitations such as restricted scale and the over-cleaning of images, which differ from real-world clinical scenarios. We present the Ganzhou Municipal Hospital Gland dataset (GZMH_GLAND), a collection of breast cancer pathological images sourced from a clinical environment, designed for gland segmentation and classification of benign and malignant tumors. The dataset comprises 1,420 high-power field images (1024×1024 pixels) extracted from 31 breast cancer patients. It is accompanied by pixel-level segmentation masks and image-level classification labels. To evaluate clinical generalization and prevent data leakage, a patient-level data splitting strategy was employed. Benchmark testing using classical segmentation and classification networks demonstrates that this dataset preserves the morphological diversity and clinical tissue artifacts of breast glands. This dataset provides a resource for developing and validating automated breast cancer diagnostic models.  breast cancer; pathological images; gland segmentation; benign and malignant classification Libingbing Liuyongan Zhengyihong Wanghuadeng Panxipeng Lanrushi Liangli PUBLIC https://creativecommons.org/licenses/by/4.0/ dataset Science Data Bank