Predicative value of radiomics nomogram based on ¹⁸F-FDG PET/CT for the prognosis of patients with postoperative gastric carcinoma

Objective: To investigate the clinical value of radiomics nomogram, which is established by ¹⁸F-fluorodeoxyglucose (FDG) PET/CT radiomics signature combined with clinical-pathologic risk factors, in predicting the prognosis of patients with postoperative gastric carcinoma. Methods: ¹⁸F-FDG PET/CT data of 207 patients (143 males, 64 females, age range: 20-85 years) with postoperative gastric carcinoma from January 2008 to August 2015 was reviewed retrospectively. Patients were divided into training group (n = 104) and validation group (n = 103), and the clinicopathologic information and disease-free survival (DFS) data were acquired. Significant textural features were selected from PET/CT images, and radiomics score (RS) for individual patient was calculated based on the radiomics signatures. The relationship between RS and DFS was analyzed. Cox regression analysis was performed to determine the risk factors offside. The radiomics nomogram, obtained from combination of RS with clinicopathologic risk factors, was established and further evaluated in predictive value for recurrence or metastasis of postoperative gastric carcinoma, and the concordance index (C-index) was calculated. Results: Cox regression analysis demonstrated that RS, tumor location, depth of invasion, lymph node metastasis, and distant metastasis were the significant risk factors for DFS (hazard ratios: 2.148-2.828, all P < 0.05). The radiomics nomogram combined with RS and 4 clinicopathologic risk factors had a better prediction for the estimated DFS, comparing to RS alone. C-index of radiomics nomogram and RS were 0.830 and 0.700 in training group, and 0.776 and 0.681 in validation group, respectively. Conclusion: Radiomics nomogram which is established by radiomics signatures and clinicopathologic risk factors may be better for predicting DFS of patients with postoperative gastric carcinoma. (authors)