TDM-STARGAN: STARGAN USING TIME DIFFERENCE MAP TO GENERATE DYNAMIC CONTRAST-ENHANCED MRI FROM ULTRAFAST DYNAMIC CONTRAST-ENHANCED MRI

Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is a sensitive imaging technique to manage many types of cancer including breast cancer. The conventional DCE-MRI takes a long time (7- 12 minutes) to acquire and there is a clinical need to reduce scan time. Ultrafast DCEMRI takes less than a minute to acquire and has sufficient information relative to conventional DCE-MRI. We propose a generative adversarial network (GAN) to generate the delay phase of synthetic conventional DCE-MRI from ultrafast DCE-MRI. We allow our model to better generate the area expected to be a lesion through the difference map of different phases to incorporate time-varying enhancement patterns. The difference map also allows us to generate pseudo tumor labels for segmentation. Our approach was trained and tested on 300 cases using three evaluation metrics. Our method showed better performance (structural similarity index map increase of 11.69%) compared to Pix2Pix baseline method.