Dynamic contrast-enhanced computed tomography in metastatic nasopharyngeal carcinoma : reproducibility analysis and observer variability of the distributed parameter model

• Main Authors: Ng, Quan-Sing, Thng, Choon Hua, Lim, Wan Teck, Hartono, Septian, Thian, Yee Liang, Lee, Puor Sherng, Tan, Daniel Shao-Weng, Tan, Eng Huat, Koh, Tong San
• Other Authors: School of Electrical and Electronic Engineering
• Format: Article
• Language: English
• Published: 2013
• Subjects: DRNTU::Engineering::Electrical and electronic engineering::Electronic systems::Biometrics
• Online Access: https://hdl.handle.net/10356/104227; http://hdl.handle.net/10220/16981
• Institution: Nanyang Technological University

Objectives: To determine the reproducibility and observer variability of distributed parameter analysis of dynamic contrast-enhanced computed tomography (DCE-CT) data in metastatic nasopharyngeal carcinoma, and to compare 2 approaches of region-of-interest (ROI) analyses. Methods: Following ethical approval and informed consent, 17 patients with nasopharyngeal carcinoma underwent paired DCE-CT examinations on a 64-detector scanner, measuring tumor blood flow (F, mL/100 mL/min), permeability surface area product (PS, mL/100 mL/min), fractional intravascular blood volume (v1, mL/100 mL), and fractional extracellular-extravascular volume (v2, mL/100 mL). Tumor parameters were derived by fitting (i) the ROI-averaged concentration-time curve, and (ii) the median value of parameters from voxel-level concentration-time curves. Measurement reproducibility and inter- and intraobserver variability were estimated using Bland-Altman statistics. Results: Mean F, PS, v1, and v2 are 44.9, 20.4, 7.1, and 34.1 for ROI analysis, and 49.0, 18.7, 6.7, and 34.0 for voxel analysis, respectively. Within-subject coefficients of variation are 38.8%, 49.5%, 54.2%, and 35.9% for ROI analysis, and 15.0%, 35.1%, 33.0%, and 21.0% for voxel analysis, respectively. Repeatability coefficients are 48.2, 28.0, 10.7, and 33.9 for ROI analysis, and 20.3, 18.2, 6.1 and 19.8 for voxel analysis, respectively. Intra- and interobserver correlation coefficient ranged from 0.94 to 0.97 and 0.90 to 0.95 for voxel analysis, and 0.73 to 0.87 and 0.72 to 0.94 for ROI analysis, respectively. Conclusion: Measurements of F and v2 appear more reproducible than PS and v1. Voxel-level analysis improves both reproducibility and observer variability compared with ROI-averaged analysis and may retain information about tumor spatial heterogeneity.