<p>Contrast administration during contrast enhanced computed tomography (CECT) image acquisition for liver tumour diagnosis causes mild to severe side effects in patients. To generate informative high contrast images from non-contrast CT (NCCT) images to facilitate diagnosis from NCCT images for liver tumour. 25 patients diagnosed with hepato cellular carcinoma (HCC), a primary liver cancer who underwent contrast enhanced computed tomography (CECT) examination were chosen as study population. Generative adversarial network (GAN) based domain translation network was proposed in this study to translate the NCCT images to high contrast images. Training and testing the proposed architecture were done in four case scenarios with NCCT and either hepatic arterial (HA) phase or portal venous (PV) phase image pairs considering full CT image and segmented liver. A total of 2652 images comprising of 884 NCCT images, 884 arterial phase images and 884 portal venous phase images taken from 25 patients were considered in this study for training and testing the GAN based domain translation network. The generated images for all case scenarios were clinically assessed in terms of tumour presence, tumour boundary delineation, enhancement presence within the tumour, heterogeneity within the tumour and tumour length. The results were statistically analyzed using t-test, Chi-Square test and Wilcoxon rank test. Tumour visibility is found to be higher in the segmented liver PV case (98.37%) with a significantly higher tumour boundary delineation. Enhancement within the tumour was the highest (88.07%) in segmented HA case whereas PV phase showed 87.41% and 68.14% for segmented and full CT cases respectively. Generated images belonging to the segmented cases of HA as well as PV phases show better clinical significance than the full CT case scenarios.</p>