ABSTRACT
Objective:To investigate the value of baseline 18F-fluorodeoxyglucose (FDG) PET/CT radiomics features in predicting the prognosis of non-small cell lung cancer (NSCLC) before treatment. Methods:From January 2016 to August 2018, a total of 300 patients with solitary NSCLC (189 males, 111 females, age (62.3±9.0) years) who underwent 18F-FDG PET/CT imaging before treatment in Fourth Hospital of Hebei Medical University were retrospectively analyzed. According to the ratio of 7∶3 (R language), 300 patients were randomly divided into training group ( n=210) and validation group ( n=90). LIFEx software package was used to extract the PET and CT radiomics features of primary focus in 300 NSCLC patients. The least absolute shrinkage and selection operator (LASSO) algorithm combined with Cox proportional hazard regression model were used to select radiomics features and clinical features for predicting overall survival (OS) and progression-free survival (PFS) in training group. Then radiomics model, clinical model and complex model which integrated the two were established and the radiomics score (Rad-score), clinical score and complex score of each patient were calculated. Data of validation group was used to validate each training model. Efficiencies of each model in predicting the prognosis of patients with NSCLC were further evaluated by the concordance index (C-index), and a nomogram was developed based on the best prediction model. Results:In training group, the C-indices of predicting OS and PFS in NSCLC patients of radiomics model were 0.762 and 0.724 respectively, which were 0.834 and 0.780 respectively in clinical model, and were 0.842 and 0.787 respectively in complex model. Cox multivariate analysis showed that both Rad-score and complex score were independent prognostic factors for OS (hazard ratio ( HR): 1.804, 9.996, 95% CI: 1.023-3.184, 4.582-21.808, both P<0.05) and PFS ( HR: 1.771, 5.627, 95% CI: 1.138-2.756, 3.429-9.234, both P<0.05). Conclusions:Pre-treatment 18F-FDG PET/CT radiomics can predict OS and PFS of NSCLC patients. The complex model based on the combination of radiomics and clinical model is effective in predicting the prognosis of NSCLC patients, and the nomogram of complex model is simple and convenient to assist clinical decision-making.