Study on the diagnostic value of dual-energy CT imaging omics characteristic model in diagnosis of liver metastasis of colorectal cancer
双能量CT影像组学特征模型对结直肠癌肝转移的诊断价值研究

摘要

目的 探讨双能量CT影像组学特征模型对结直肠癌肝转移的诊断价值研究。方法 选择我院于2021年1月-2024年12月期间内收治的结直肠癌患者160例，其中肝转移组80例、非转移组80例。通过双能量CT扫描获取100 kVp、Sn 140 kVp、融合图像及最高对比度噪声比（CNR）图像，并进行比较。提取影像组学特征并构建支持向量机（SVM）、随机森林（RF）和逻辑回归（LR）模型，评估模型在训练集（70%）和验证集（30%）中的诊断效能。结果 100 kVp图像、Sn 140 kVp图像、融合图像及CNR图像各序列，同时性肝转移灶CT值比较差异无统计学意义（P＞0.05）。图像噪声、同时性肝转移灶CT差值及CNR 比较均有统计学意义（均P＜0.05）。SVM模型在验证集中表现最优，准确度为0.944、AUC为0.982、灵敏度为0.929、特异度为0.958。结论 双能量CT影像组学特征构建的SVM模型可显著提升结直肠癌肝转移的诊断准确性，有较高应用价值。

Abstract

Objective Exploring the diagnostic value of dual energy CT imaging omics feature model for liver metastasis of colorectal cancer.
Methods Select 160 colorectal cancer patients admitted to our hospital from January 2021 to December 2024, including 80 cases in the liver metastasis group and 80 cases in the non metastasis group. Obtain 100 kVp, Sn 140 kVp, fused images, and highest contrast to noise ratio (CNR) images through dual energy CT scanning, and compare them. Extract radiomics features and construct support vector machine (SVM), random forest (RF), and logistic regression (LR) models to evaluate the diagnostic performance of the models in the training set (70%) and validation set (30%).
Results There was no statistically significant difference (P>0.05) in the CT values of synchronous liver metastases among the sequences of 100 kVp images, Sn 140 kVp images, fused images, and CNR images. There were statistically significant differences in image noise, CT differences of synchronous liver metastases, and CNR (all P<0.05). The SVM model performed the best in the validation set, with an accuracy of 0.944, AUC of 0.982, sensitivity of 0.929, and specificity of 0.958.
Conclusion   The SVM model constructed based on dual energy CT imaging omics features can significantly improve the diagnostic accuracy of colorectal cancer liver metastasis and has high application value.