| Gastric cancer is a common malignancy and has a high recurrence rate and poor prognosis.In clinical decision making,methods based on TNM staging system cannot effectively predict patient prognosis,and even for patients with the same TNM stage,the prognosis varies greatly.Gastric cancer clinics urgently need more accurate prognostic prediction methods to guide clinical decision making.The tumor microenvironment is a complex ecosystem consisting of various types of immune cells,fibroblasts,extracellular matrix and blood vessels.The immune,stromal components and blood vessel formation of the tumor microenvironment all interact closely with cancer cells and play key roles in cancer progression,metastasis,treatment response and drug resistance.Numerous studies have confirmed the important prognostic predictive value of the tumor microenvironment in a variety of tumors,including gastric cancer.Firstly,tumor stromal microenvironment is strongly correlated with prognosis.The aim of this study was to develop a non-invasive radiomics model to assess tumor stroma type and prognosis.The study retrospectively analyzed data of 2209 patients from two centers.α-SMA and POSTON were used as indicators to detect tumor stroma,and gastric cancer stroma was classified into four types based on the immunohistochemical staining results of α-SMA and POSTON.A deep learning model for predicting stroma type of gastric cancer based on preoperative CT images was trained with validation cohort 1 AUC 0.96-0.98 and external validation cohort 1 AUC 0.89-0.94.The model was significantly associated with disease-free survival and overall survival(p<0.0001).Then,we evaluated the relationship between the model and the benefits of chemotherapy in patients with stage Ⅱ or Ⅲ gastric cancer.In predicted stromal types 1 and 2 subgroups,survival was improved in patients who received chemotherapy compared with those who did not(stage Ⅱ HR:0.48,95%CI 0.29-0.77,p=0.0021,stage Ⅲ HR 0.70,95%CI 0.57-0.85,p=0.00042).However,in the predicted stromal types 3 and 4 subgroups,adjuvant chemotherapy was not associated with survival(HR 1.48,95%CI 1.08-2.03,p=0.013).Thus,the deep learning model achieves an accurate assessment of the tumor stroma is an independent predictor of prognosis in gastric cancer,and is associated with chemotherapy benefit.Secondly,angiogenesis in the microenvironment was highly correlated with tumor prognosis.Microvessel density(MVD)is a common index for quantifying tumor angiogenesis,which is a prognostic indicator for many cancers.The aim of this study was to develop a non-invasive radiomics approach for MVD.The study retrospectively analyzed the date of 699 patients with surgically resected gastric cancer from two cancer centers.MVD was determined by immunohistochemistry using CD34.Then,training and validation of the radiomics score for assessing MVD were performed.The AUC of the radiomics score was 0.735(95%CI,0.669-0.801),0.751(0.671-0.831)and 0.736(0.629-0.843)in the training cohort and validation cohort,respectively.In both the training and validation cohorts,radiomics score were significantly associated with disease-free survival and overall survival(HR range:1.644-1.931,all P<0.05).Thus,radiomics score is a reliable model for individualized assessment of MVD in gastric cancer.In summary,this study provided new ideas for tumor prognosis prediction.And,our research provide new tools for personalized diagnosis and treatment of gastric cancer. |
PDF Full Text Request