Integrating Imaging, Clinical, and Pathological Data on Post-hepatectomy Liver Failure

We present a multi-institutional dataset of preoperative Gd-EOB-DTPA-enhanced MRI scans, comprising 14,895 images across three academic medical centers. This comprehensive dataset includes more than 20,000 expert annotations of key anatomical structures - such as the liver, Couinaud liver segments, liver tumors, spleen, and psoas muscles - enabling automated quantification of future liver remnant (FLR) volume and functional assessment. Additionally, it incorporates detailed clinicopathological variables, with all annotations subjected to a two-tier radiological review process to ensure labeling accuracy and reliability. By integrating FLR volume, Gd-EOB-DTPA-enhanced MRI-derived functional parameters, and clinical risk factors, this dataset serves as a foundational resource for developing interpretable and generalizable AI models.本研究构建了一套多中心数据集，涵盖来自三家学术医疗中心的共14895幅术前钆塞酸二钠（Gd-EOB-DTPA）增强磁共振成像（MRI）扫描图像。该数据集包含超过2万条针对肝脏、库伊纳尔德肝段（Couinaud liver segments）、肝脏肿瘤、脾脏及腰大肌等关键解剖结构的专家标注，可实现对未来肝剩余体积（future liver remnant，FLR）的自动化定量分析与功能评估。此外，该数据集还纳入了详细的临床病理变量，且所有标注均经过两级放射学审核流程，以确保标注的准确性与可靠性。本数据集整合了FLR体积、Gd-EOB-DTPA增强MRI衍生的功能参数以及临床危险因素，可作为开发可解释、可泛化人工智能（Artificial Intelligence，AI）模型的基础资源。