Quantitative analysis of CT radiomics features of nasopharyngeal carcinoma(PDF)
《中国医学物理学杂志》[ISSN:1005-202X/CN:44-1351/R]
- Issue:
- 2019年第5期
- Page:
- 551-555
- Research Field:
- 医学影像物理
- Publishing date:
Info
- Title:
- Quantitative analysis of CT radiomics features of nasopharyngeal carcinoma
- Author(s):
- LI Tengxiang1; GONG Guanzhong2; QIU Qingtao2; YIN Yong2; QIU Xiaoping1
- 1. School of Nuclear Science and Technology, University of South China, Hengyang 421001, China; 2. Department of Radiophysical Technology, Shandong Cancer Hospital, Shandong University, Ji’nan 250117, China
- Keywords:
- Keywords: nasopharyngeal carcinoma; radiomics; quantitative analysis; heterogeneity
- PACS:
- R318；R739.6
- DOI:
- DOI:10.3969/j.issn.1005-202X.2019.05.011
- Abstract:
- Abstract: Objective To analyze the differences and change patterns of radiomics features in the multi-phase contrast-enhanced computed tomography (CT) images of nasopharyngeal carcinoma (NPC). Methods The CT images of 41 patients with newly diagnosed NPC were retrospectively analyzed. Each group of CT images consisted of 3 phases, namely plain scan, arterial phase and venous phase. The gross tumor volume (GTV) and pterygoid muscle at each phase were set as regions of interest. Finally, the differences in CT radiomic features of GTV and pterygoid muscle were compared among different phases. Results The comparisons of plain scan and arterial phase, plain scan and venous phase, arterial phase and venous phase showed that there were 18, 21, 9 and 31, 31, 22 statistically significant features of GTV and pterygoid muscle, respectively. The comparison among arterial phase, venous phase and plain scan revealed that the interquartile range of intensity direct in GTV was increased with the increase of the quantile. The statistically significant features of GTV and pterygoid muscle which were found in the comparison among plain scan, arterial phase and venous phase were mainly distributed in the gray level co-occurrence matrix. Among them, 18 common features of GTV and pterygoid muscle were analyzed with receiver operating characteristic curve. There were 10 feature predicting that GTV had an AUC≥0.9 (Area Under Curve), and all of these features were distributed in the gray level co-occurrence matrix and intensity direct. Conclusion CT radiomics can quantitatively analyze the subtle discrepancy of NPC and normal tissues in multi-phase CT images, reflect the dynamic variations and provide a reference for the accurate diagnosis and quantitative analysis of NPC.
Last Update: 2019-05-23