ENHANCEMENT OF BRAIN MRI IMAGE DATABASE USING CYCLEGAN

Abstract

Generating tumor images on brain MRI scans at random locations can assist medical researchers and medical students in predicting the likelihood of tumors. However, brain MRI images with tumors are rare in practice, making the collection of MRI image data with brain tumors a time-consuming process. In this study, we propose the application of CycleGAN to generate T2 pulse sequence MRI images of the human brain from T2 Flair pulse sequence images of the same type and vice versa, thereby increasing the number of MRI images of various types. The returned results will be evaluated and compared with other studies based on FID, SSMI, and PSNR metrics.