| 杨超帆,梁长华,梁盼,吕东博,甄思雨,潘犇,杨鑫淼,王慧慧.肿瘤-脾脏CT影像组学联合临床及CT表现预测进展期胃癌Lauren类型[J].中国医学影像技术,2025,41(12):2016~2021 |
| 肿瘤-脾脏CT影像组学联合临床及CT表现预测进展期胃癌Lauren类型 |
| Tumor-spleen CT radiomics combined with clinical and CT findings for predicting Lauren classification of advanced gastric cancer |
| 投稿时间：2025-01-14 修订日期：2025-11-09 |
| DOI：10.13929/j.issn.1003-3289.2025.12.017 |
| 中文关键词: 胃肿瘤 脾 体层摄影术,X线计算机 影像组学 Lauren分型 |
| 英文关键词:stomach neoplasms spleen tomography, X-ray computed radiomics Lauren classification |
| 基金项目:河南省慢病防治与智慧健康管理重点实验室开放基金课题(ZYYC2024MB)。 |
|
| 摘要点击次数: 199 |
| 全文下载次数: 38 |
| 中文摘要: |
| 目的 探讨肿瘤-脾脏CT影像组学联合临床及CT表现预测进展期胃癌Lauren类型的价值。方法 回顾性纳入331例进展期胃癌(Lauren类型为137例肠型、194例非肠型),按8∶2比例划分训练集(n=265,含110例肠型及155例非肠型)与测试集(n=66,含27例肠型及39例非肠型)。以单因素和多因素logistic回归分析不同Lauren类型进展期胃癌的临床资料及CT表现,筛选其独立预测因素。基于增强静脉期CT提取及筛选肿瘤、脾脏及肿瘤-脾脏影像组学特征,分别构建影像组学模型,并以后者联合临床及CT相关独立预测因素构建联合模型;评估各模型预测进展期胃癌Lauren类型的效能及其临床净收益。结果 Borrmann类型、有无淋巴结转移及脾脏静脉期CT值标准差均为非肠型进展期胃癌的独立预测因素(P均＜0.05)。肿瘤、脾脏、肿瘤-脾脏影像组学模型及联合模型在训练集的曲线下面积(AUC)分别为0.771、0.752、0.814及0.862,在测试集分别为0.743、0.748、0.774及0.859。联合模型在训练集的AUC均高于其他模型(P均＜0.05),在测试集高于脾脏影像组学模型(P＜0.05),其在训练集的净收益亦高于其他模型;阈值取0.45~0.90时,联合模型在测试集的净收益高于其他模型。结论 基于肿瘤-脾脏影像组学联合临床及CT表现能有效预测进展期胃癌Lauren类型。 |
| 英文摘要: |
| Objective To explore the value of tumor-spleen CT radiomics combined with clinical and CT findings for predicting Lauren classification of advanced gastric cancer. Methods Totally 331 patients with advanced gastric cancer (137 cases of intestinal type and 194 cases of non-intestinal type of Lauren classification) were retrospectively enrolled and divided into training set (n=265, including 110 intestinal type and 155 non-intestinal type) and test set (n=66, including 27 intestinal type and 39 non-intestinal type) at the ratio of 8 ∶ 2. Univariate and multivariate logistic regression analyses were performed on clinical and CT data to obtain the independent predictive factors of advanced gastric cancer of different Lauren classification. Radiomics features of tumor, spleen and tumor-spleen were extracted and screened from enhanced venous-phase CT to construct the corresponding radiomics models, and a combined model was built based on radiomics features of tumor-spleen combined with clinical and CT-related independent predictive factors . The performance of the models for predicting Lauren classification of advanced gastric cancer were evaluated, and their clinical net benefit were assessed. Results Borrmann classification, with or without lymph node metastasis and enhanced venous-phase standard deviation of CT value of spleen were all independent predictive factors of non-intestinal advanced gastric cancer (all P＜0.05). The area under the curve (AUC) of tumor radiomics, spleen radiomics, tumor-spleen radiomics and combined models was 0.771, 0.752, 0.814 and 0.862 in training set, 0.743, 0.748, 0.774 and 0.859 in test set, respectively. The combined model had higher AUC in training set than other models, higher AUC than spleen radiomics model in test set (all P＜0.05), also higher clinical net benefit than other models in training set. Taken threshold of 0.45 to 0.90, the clinical net benefit of combined model was higher than that of other models in test set. Conclusion Tumor-spleen CT radiomics combined with clinical and CT features could effectively predict Lauren classification of advanced gastric cancer. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
