The pseudo-siamese framework combines Transformer and CNN for medical image generation.

Multi-phase medical imaging can provide significant improvement in disease multi-modal diagnosis. However, medical image data often suffer from modality missing issues. Therefore, synthesizing missing phases using available phases is of great clinical significance. Existing generation methods often focus on either local lesion details or global structural features, while medical image generation requires both aspects. To address this, we propose the siam TC-GAN, a Generative Adversarial Network (GAN) based on pseudo-siamese architecture, which can extract both global and local information from images and perform multi-scale feature fusion. In addition, we propose a novel HE-loss that guides the generator towards more realistic images from a grayscale feature perspective. Experimental results demonstrate the effectiveness of our proposed method on contrast-enhanced CT and contrast-enhanced MRI.