Texture analysis and metabolic parameters of 18F-FDG PET/CT to predict primary tumour response and prognosis of paediatric soft tissue sarcomas.

Abstract

INTRODUCTION We aimed to investigate the value of primary tumour F-18 fluorodeoxyglucose (18F-FDG) parameters and textural features in predicting tumour response to neoadjuvant chemoradiotherapy (neo-CRT) and prognosis in paediatric patients with soft tissue sarcoma (STS). MATERIALS AND METHODS Twenty-eight paediatric patients with STS who underwent 18F-FDG PET/CT studies before neo-CRT were included in this retrospective and single-center study. SUVmax, SUVpeak, SUVmean, metabolic tumour volume (MTV, 40% SUVmax), total lesion glycolysis (TLG), and textural features were extracted from the primary tumour volumes delineated semiautomatically on the baseline PET images. Patients were classified as responders or non-responders according to Response Evaluation Criteria in Solid Tumors 1.1. A receiver operating characteristic (ROC) analysis was performed. The highest AUC values within their respective quantitative groups were selected for further analysis, including logistic regression analysis for response prediction and Cox regression analysis for survival prediction. RESULTS In univariate analysis SUVmax > 13.0 (p = 0.009), SUVpeak > 12.7 (p = 0.017), Histogram Entropy > 0.97 (p = 0.036), and NGTDM Busyness < 0.37 (p = 0.005) were associated with tumour response for the median follow-up of 25 months. NGTDM Busyness was an independent predictor for the treatment response (OR: 30.5; 95% CI: 1.50-618.5; p = 0.026). Age was associated with progression (Cut-off: 11 years, [AUC:0.73 (95% CI: 0,53 - 0,93)] 𝑝=0.022). Progression-free survival outcomes were assessed in aged > 11 years subpopulation. PFS was significantly shorter in patients with high GLSZM_GLNU (p = 0,024), GLSZM_ZSNU (p = 0,003), and TLG (p = 0,016). In multivariate analysis GLSZM_ZSNU > 13,04 (HR: 11.61; 95% CI: 1.35-54.02; p = 0.026) was an independent predictor of PFS in subpopulation aged > 11 years. CONCLUSION Heterogeneity texture features Histogram Entropy and NGTDM Busyness and metabolic PET parameters (SUV max and SUVpeak) can predict tumour response. In aged > 11 years patients subgroup analyses, GLSZM ZSNU was an independent factor for PFS.