Accelerating cardiac DTI using 3D SwinGAN-based multi-directional joint reconstruction

Cardiac Diffusion Tensor Imaging (cDTI) is a non-invasive imaging technique that allow us to accurately understand myocardial fiber structure, which is important for analyzing the causes and early diagnosis of cardiovascular diseases. However, it remains challenging to use cDTI in clinical settings due to its scanning time and motion sensitivity. In this paper, we present a multi-directional joint reconstruction method for cardiac magnetic resonance diffusion-weighted images, utilizing a 3D Swin Transformer within the SwinUnet framework, and introduce a generative adversarial network (GAN) to improve the image reconstruction quality and accuracy. Experimental results on diffusion-weighted images of human isolated hearts demonstrate that the proposed method exhibits lower reconstruction errors in both visual and quantitative DTI metrics compared to existing CNN-based 3D Unet image reconstruction techniques.