UCS-MedTranGAN: Unsupervised Medical Image Transformation Using CSGAN
S. Poonkodi, M. Kanchana · 2022 4th International Conference on Smart Systems and Inventive Technology (ICSSIT) · 2022
Medical Image transformation is an emerging area in computer vision technology with several applications within the medical field. Supervisedlearning requires pixel-wisepairedco-registered image datasets to perform the image transformation. In realistic situations, gathering these types of datasets is a challenging task. And the unsupervisedimage transformation architecture resulted in the blurred transformed images with non-realistic data. A novel unsupervised medical image transformation model named as UCS -MedTranGAN is proposed. The proposed model uses a new non-adversarial cyclic synthesized losses which reduces the content and perceptual loss in the transformed medical images. This work is compared with other unsupervised transformation architecture which shows the implementation of the proposed model based on the two tasks i) the transformation of Positron Emission Tomography image to Computed Tomography image and ii) rectification of MR movement.