基于四维计算机断层扫描图像的非小细胞肺癌影像组学特征稳定性分析

Stability of Radiomics Features of Non-Small Cell Lung Cancer Based on Four-Dimensional Computed Tomography

目的基于四维计算机断层扫描图像(4DCT)分析非小细胞肺癌影像组学特征的稳定性,评估该方法用于影像组学特征筛选的可行性。资料与方法选取20例非小细胞肺癌患者共200组CT影像,提取不同时相的影像组学特征,包括常规及纹理特征共60项。统计各特征在不同时相间的一致性相关系数(CCC)。引入动态范围指标(DR)量化各特征在不同患者间的差异。结果CCC>0.85的特征占总数的70%(42/60);CCC>0.95的特征占总数的35%(21/60)。DR>0.85的特征占总数的91%(55/60);DR>0.95的特征占总数的21%(13/60)。阈值在0.9~0.95范围间的筛选差异显著。使用CCC>0.95、DR>0.95筛选出10项特征,占本研究总特征数的16.7%(10/60)。结论4DCT多时相对比能够实现非小细胞肺癌影像组学特征稳定性分析,可作为影像组学特征筛选的一种方法。

Purpose To analyze the feasibility of filtering radiomics features of non-small cell lung cancer based on four-dimensional computed tomography(4DCT).Materials and Methods Twenty patients with non-small cell lung cancer were retrospectively selected.The features of different time-phase images were extracted.There were 60 regular and texture features.The concordance correlation coefficient(CCC)of each feature in different time phase images was statistically introduced.The dynamic range(DR)quantifies the differences between features in different patients.Results The features with a CCC>0.85 accounted for 70%(42/60)of the total,compared to 35%(21/60)of total which had a CCC>0.95.91%(55/60)of the features had a DR>0.85,compared with 21%(13/60)of the features had a DR>0.95.There was a significant filtering difference in threshold range between 0.9 and 0.95.Ten features were filtered using CCC>0.95 and DR>0.95,accounting for 16.7%(10/60)of the total number of features in the study.Conclusion 4DCT image can achieve the stability analysis of radiomics features non-small cell lung cancer.It can be used as a method of filtering for radiomics features.