Please click here to download the model.
CTNomogram4AGC-N: A CT based deep learning radiomic nomogram (DLRN) for preoperative N staging in patients with advanced gastric cancer was developed and validated on multi-center datasets. The DLRN has good predictive ability for preoperatively discriminating pathologic N0, N1, N2, N3a, and N3b, i.e. the accurate categorization. We suggest the clinicians and physicians to use it to supplement the clinical judgment. To use this model, one should outline the 2D tumor region in the images of arterial phase CT, venous phase CT, and unenhanced CT. Three radiomic signatures could be obtained by combining 19 radiomic features, including deep learning features and hand-crafted features. Two signatures and the CT-finding N staging are input into the final DLRN. Then the possible exact N stage can be given.
Abstract :
Preoperative evaluation of the number of lymph node metastasis (LNM) is the basis of individual treatment of advanced gastric cancer (AGC). However, the routinely used preoperative determination method is not accurate enough. We develop a deep leaning radiomic nomogram (DLRN) based on the multi-phase CT images for preoperatively determining the number of LNM in AGC. The DLRN shows a good discriminability of the number of LNM on AGC cohorts from 6 Chinese centers (n=679), which is significantly superior to the routinely used clinical N stages, tumor size, and clinical model. In addition, the DLRN also shows good performance on an Italian cohort (n=51). Besides, the DLRN is significantly associated with overall survival of AGC patients (n=271). More interestingly, the cross-cancer analysis show that the DLRN can well discriminate LNM in colerectal cancer (n=80). In staging-oriented treatment of gastric cancer, this preoperative nomogram could provide baseline information for individual treatment of AGC.