LH2-data-labs/multimodal-oncology-atlas

LH2数据—多模态肿瘤数据集是一个大规模、多模态的肿瘤数据集，其独特之处在于以非高加索人种患者群体为核心。该数据集解决了当前肿瘤数据集主要来自西方高加索人种的局限性，提供了涵盖基因组、临床、影像和生物样本的多模态资源，覆盖了全球南部的15个国家。样本包含20名匿名癌症患者的结构化临床记录，完整数据集包括约25,000名患者（目标在12个月内达到50,000名）。数据集包含全外显子测序（WES）、数字病理（WSI）、临床记录和生物样本库等多种数据模态，所有数据均以患者为单位进行链接。临床记录包含28个字段，涵盖癌症类型、分期、治疗反应等多个方面。The LH2 Data — Multimodal Oncology Dataset is a large-scale, multimodal oncology dataset built around a principle rare in the field: placing non-Caucasian patient populations at the centre. It addresses the well-documented limitation of oncology datasets being overwhelmingly drawn from Caucasian, Western populations by offering a deeply multimodal oncology resource — genomic, clinical, imaging, and biospecimen — anchored across the Global South. The sample contains structured clinical records for 20 de-identified cancer patients, with the complete dataset including ~25,000 enrolled patients (target: 50,000 within 12 months) across 15 countries. The dataset includes whole exome sequencing (WES), digital pathology (WSI), clinical records, and biobank specimens, all linked at the patient level. The clinical record schema comprises 28 fields covering cancer type, staging, treatment response, and more.