Program Information

Repeatability Analysis of GTV Texture Feature Extracted From Four-Dimensional CT in NSCLC

J Duan

Shandong Cancer Hospital Affiliated to Shandong University, Jinan, Shandong province

Presentations

SU-I-GPD-J-83 (Sunday, July 30, 2017) 3:00 PM - 6:00 PM Room: Exhibit Hall

Purpose: Objectively and reproducibly quantify various imaging features is important to reveal the underlying biology of cancer. In this study, we assessed the repeatability of GTV texture feature extracted from four-dimensional CT (4D-CT) in NSCLC patients.

Methods: Five NSCLC patients scanned by 4D-CT were consecutively enrolled in this research. For each patient, GTV was delineated manually in each phase of the 10 four-dimensional CT phases and agreed by two observers. 9 kinds of texture features, including angular second moment, inertia, inverse difference moment, entropy, correlation, sum average, difference average, sum entropy, difference entropy, were extracted from the GTV for each phase based on gray level co-occurrence matrix. Interclass Correlation Coefficient (ICC) analysis were performed to quantify the reproducibility of a feature. The ICC value of ≥ 0.85 was considered a reproducible feature.

Results: The features were found to have little dependency on the phases of 4D-CT. Most texture features were found to have significantly higher reproducibility (ICC ≥ 0.89, p<0.001) except sum average (ICC = 0.75±0.21, p>0.001).

Conclusion: The study shows most texture features are repeatable between different 4D-CT phases in NSCLC. However, the reproducibly of radiomics features (including texture features) causing by different reconstruction algorithms and the slice thickness, should be further study to reveal the underlying biology of cancer.


Contact Email: