Cardiac vessel segmentation from angiographic images is of great importance for diagnosing and treating cardiovascular disease. Despite the great value that segmentations provide achieving high-quality annotated datasets of cardiac vessels is difficult and costly to achieve due to the complexity in time and manual labor required. This work proposes an advanced conditional generative adversarial network (cGAN) framework using hybrid conditioning using edge maps and distance maps to help provide complementary structural cues for the learning of vessel orientation. The edge map provides a boundary while the distance maps provide richer contextual information on the thickness and continuity of the vessel structure providing this framework additional prior information to ensure that the generated masks are accurate. The cGAN framework uses a generator based on residual embedding network architecture while the discriminator uses a patch based discriminator. The generator and discriminator are trained with a loss based on adversarial learning as well as L1 loss. It is the evidence from the experiments outlined in this work, that support the evidence that hybrid conditioning, outperformed edge conditioning, with better Dice coefficient, IoU, precision, recall and pixel-wise accuracy in limited annotated data. Overall, this study provides evidence that hybrid conditioning by using multiple structural cues referencing actual cardiac vessels improves quality and robustness of vessel segmentation in sparse real-world data.