透明细胞肾细胞癌（ccRCC）是影响肾脏的最常见恶性肿瘤，在转移性病例中表现出较差的预后。阐明 ccRCC 的组成有望发现高度敏感的生物标志物。我们的目标是利用栖息地成像技术并整合多模态数据来精确预测转移风险，最终实现早期干预并提高患者生存率。对 2013 年 4 月至 2013 年 4 月至 2013 年 3 家医院的 263 名 ccRCC 患者进行了回顾性分析。 2021年3月，综合分析术前CT图像、超声图像和临床数据。来自山东大学齐鲁医院两个院区的患者被分配到训练数据集，而第三医院作为独立测试数据集。采用稳健的共识聚类方法，利用对比增强 CT 图像将原发肿瘤空间分类为不同的子区域（即栖息地）。从这些肿瘤亚区域中提取放射组学特征，然后进行简化，以确定有意义的特征，用于构建 ccRCC 转移风险评估的预测模型。此外，通过整合超声图像特征和临床数据来构建和比较替代模型，探索放射组学在预测 ccRCC 转移风险方面的潜在价值。在本研究中，我们在肿瘤区域内进行 k 均值聚类，以生成三个不同的肿瘤亚区域。我们量化了每个次区域的豪恩菲尔德单位 (HU) 值、体积分数以及高风险组和低风险组的分布。我们的调查重点关注 Habitat1 和 Habitat3 的 252 名患者，以评估这两个分区的区分力。然后，我们根据从 CT 和超声图像以及临床数据中提取的放射组学特征，开发了 ccRCC 转移风险分类的风险预测模型。组合模型和 CT_Habitat3 模型在训练数据集中的 AUC 值分别为 0.935 [95%CI: 0.902-0.968] 和 0.934 [95%CI: 0.902-0.966]，而在独立测试数据集中，它们实现了 AUC 值值分别为 0.891 [95%CI: 0.794-0.988] 和 0.903 [95%CI: 0.819-0.987]。我们已经确定了一种非侵入性成像预测因子，并且所提出的次区域放射组学模型可以准确预测转移风险在ccRCC。该预测工具具有临床应用潜力，可为 ccRCC 患者完善个体化治疗策略。版权所有 © 2024 大学放射科医生协会。由爱思唯尔公司出版。保留所有权利。

Clear cell renal cell carcinoma (ccRCC) is the most common malignant neoplasm affecting the kidney, exhibiting a dismal prognosis in metastatic instances. Elucidating the composition of ccRCC holds promise for the discovery of highly sensitive biomarkers. Our objective was to utilize habitat imaging techniques and integrate multimodal data to precisely predict the risk of metastasis, ultimately enabling early intervention and enhancing patient survival rates.A retrospective analysis was performed on a cohort of 263 patients with ccRCC from three hospitals between April 2013 and March 2021. Preoperative CT images, ultrasound images, and clinical data were comprehensively analyzed. Patients from two campuses of Qilu Hospital of Shandong University were assigned to the training dataset, while the third hospital served as the independent testing dataset. A robust consensus clustering method was used to classify the primary tumor space into distinct sub-regions (i.e., habitats) using contrast-enhanced CT images. Radiomic features were extracted from these tumor sub-regions and subsequently reduced to identify meaningful features for constructing a predictive model for ccRCC metastasis risk assessment. In addition, the potential value of radiomics in predicting ccRCC metastasis risk was explored by integrating ultrasound image features and clinical data to construct and compare alternative models.In this study, we performed k-means clustering within the tumor region to generate three distinct tumor subregions. We quantified the Hounsfiled Unit (HU) value, volume fraction, and distribution of high- and low-risk groups in each subregion. Our investigation focused on 252 patients with Habitat1 + Habitat3 to assess the discriminative power of these two subregions. We then developed a risk prediction model for ccRCC metastasis risk classification based on radiomic features extracted from CT and ultrasound images, and clinical data. The Combined model and the CT_Habitat3 model showed AUC values of 0.935 [95%CI: 0.902-0.968] and 0.934 [95%CI: 0.902-0.966], respectively, in the training dataset, while in the independent testing dataset, they achieved AUC values of 0.891 [95%CI: 0.794-0.988] and 0.903 [95%CI: 0.819-0.987], respectively.We have identified a non-invasive imaging predictor and the proposed sub-regional radiomics model can accurately predict the risk of metastasis in ccRCC. This predictive tool has potential for clinical application to refine individualized treatment strategies for patients with ccRCC.Copyright © 2024 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved.