Abstract
1388
Objectives Impacts of respiratory motion on PET texture features had not been fully elucidated in patients of non-small-cell lung cancer (NSCLC). Respiration-averaged CT (ACT) is clinically useful for reducing respiratory artefacts. Herein, we evaluated the impact of ACT on texture parameters in this study.
Methods We retrospectively enrolled NSCLC patients who performed 18F-FDG PET/CT for staging. Patients of M1 disease were excluded. ACT was done after PET/CT for attenuation correction and adaptive threshold for tumour segmentation was used. Texture indices were derived in tumours based on normalized grey level co-occurrence matrix (NGLCM), neighbourhood gray-tone difference matrix (NGTDM) and grey level size zone matrix (GLSZM) in both PET attenuation correction by CT (PET/CT) and ACT (PET/ACT). For PET/CT and PET/ACT parameters, variables were analyzed by Wilcoxon signed-rank test. The reproducibility and reliability were accessed by the Bland-Altman method and intraclass correlation coefficients (ICC), respectively. Receiver operating characteristic (ROC) curves, univariate and multivariate Cox regression analyses were used to identify parameters associated with disease specific survival (DSS).
Results We enrolled 56 patients. Significant higher maximum, mean, and variance of standard uptake value (SUV) and total lesion glycolysis (TLG) were found in PET/ACT. But, all texture features of NGLCM, NGTDM, and most of GLSZM revealed no different, even for the lower lung field or larger tumours. Yet, higher variations and worse correlation were noted in GLSZM indices. Multivariate analysis demonstrated uniformity (from NGLCM), coarseness (from NGTDM) and size-zone variability (from GLSZM) were significantly associated with DSS only in PET/ACT. Only entropy, which had lower variation (9.8%), showed predictive significance in both PET/CT and PET/ACT.
Conclusions Impacts of respiratory motion should be considered carefully for PET texture parameters in NSCLC especially for survival analysis.