Research on Chest Medical Image Modality Transfer Using GANs
Panlu You, Wanting Jing, Xiaolian Gao, Dapeng Cheng · 2023
MRI plays a crucial role in clinical diagnosis and lesion analysis. MR images from different sequences provide richer information, aiding healthcare professionals in making accurate clinical diagnoses. This paper proposes a self-supervised Generative Adversarial Network (SC-GAN) framework, which synthesizes Apparent Diffusion Coefficient (ADC) images from T2 Turbo Inversion Recovery Magnitude (T2 TIRM) images. The SC-GAN model introduced in this paper does not require a paired dataset, instead utilizing self-features as supervisory information to synthesize high-quality MR images. Experimental results indicate that, compared to existing GAN-based methods, this approach achieves higher levels of perceptual quality and tissue detail in the generated chest T2 TIRM images.