In this paper, we propose an interpretable deep learning-based method for diagnosis of helicobacter pylori (H. pylori) infection using double-contrast upper gastric barium X-ray images. Based on a transfer learning strategy, a deep convolutional neural network (DCNN) model, named Inception-ResNet-v2, was trained and tested to classify gastric X-ray images into two classes: infected and non-infected. In addition, an visualization technique was utilized to generate a saliency map that can indicates which anatomic regions are important for predicting the H. pylori infection. The experimental results demonstrated that the proposed method can achieve a high sensitivity and specificity in diagnosis of H. pylori infection. As a computer-aided diagnosis (CAD) system, the proposed method is also capable of providing an interpretable diagnosis to explain the relation between the image features and the prediction result.