Radiomics signature for prediction of long-term survival and recurrence patterns in patients with gastric cancer after radical gastrectomy: A multicenter study

Abstract

Background: This study aimed to develop and validate a radiomics score to predict the long-term survival and patterns of recurrence of gastric cancer (GC). Methods: A total of 513 patients who underwent radical gastrectomy for GC after curative resection between 2008 and 2016 at two institutions were analyzed. A radiomics score was generated using the least absolute shrinkage and selection operator Cox regression model on 327 patients and was validated in 186 patients. A nomogram consisting of the radiomics score and clinicopathological factors was created and compared with the tumor-lymph node-metastasis (TNM) staging system. Model performance was assessed using calibration, discrimination, and clinical usefulness. Results: The radiomics score was established based on five selected features. A higher score was significantly associated with poorer recurrence-free survival (RFS) and overall survival (OS) rates, both in the training and validation cohorts ( P < 0.05). Multivariate analysis demonstrated that the radiomics score was an independent prognostic factor for both RFS and OS ( P < 0.05). A nomogram incorporating the radiomics score had a significantly better prognostic value than the TNM system alone. Moreover, a high score was significantly associated with an increased risk of distant recurrence, a medium score was significantly associated with an increased risk of peritoneal recurrence, and a low score was significantly associated with an increased risk of locoregional recurrence, in the entire cohort ( P < 0.05). Conclusions: The newly proposed radiomics score may be a powerful predictor of long-term outcomes and recurrence patterns of GC. Further studies are warranted to confirm these findings.

Ultrasound radiomics signature for predicting central lymph node metastasis in clinically node-negative papillary thyroid microcarcinoma

“…Before the feature selection, the values of the extracted features were standardized with z scores. A three-step procedure was performed to select the robust radiomics features in the training cohort [ 14 ]. First, a univariable logistic regression analysis was performed to identify significant CLNM predictors with P < 0.05.…”