Radiomics analysis of multiphase contrast-enhanced CT imaging for differentiating primary gastric lymphoma and Borrmann type IV gastric cancer

Abstract

Abstract Clinical diagnosis of Borrmann type IV GC is highly challenging since it has many similar clinicopathologic features with PGL. Radiomics is a promising medical application that demonstrates image analysis may be improved by extracting and selecting a lot of imformation. Previous studies used texture-based or radiomics techniques to differentiate the two gastric malignancies. Nevertheless, radiomics features were extracted only from venous CT images rather than multiphase contrast-enhanced CT images. In the present study, Radiomics features were extracted from unenhanced phase, arterial phase, venous phase, double-phase(arterial + venous), three-phase (unenhanced + arterial + venous) CT images. As a result, we found Clinical model comprised of high-enhanced serosa sign, normalized CT value on venous phase, and perigastric fat infiltration showed good performance with an area under the ROC curve (AUC) of 0.902 and 0.878 in the training set and testing set. Compared with the other radiomics models, three-phase radiomics model showed better performance with AUC of 0.871 and 0.865 in the training set and testing set, respectively. The clinical-radiomics model further improved the discriminatory performance (AUC, 0.960) in the training set and testing set (AUC, 0.932). The DCA showed the highest benefit for clinical-radiomics model. Overall, this study shows that the combined clinical-radiomic model incorporating three-phase radiomics signatures and CT findings achieved a satisfactory performance in distinguishing PGL from Borrmann type IV GC.