Peripheral artery disease is an atherosclerotic disease that represents accumulation of cholesterol in peripheral arteries, causing thickening of the arterial walls making it difficult for the blood to flow to lower extremities. For these reasons, it is extremely important to detect the disease as soon as possible. In this paper we provide a deep learning approach for lumen and intima segmentation of porcine femoral arteries as an important pre-step for the use in humans. Proposed methods focus on different convolutional neural networks and highlights the difference in results and include modified U-Net, SegNet and PSPNet networks. Methods are evaluated on manually annotated OCT images in terms of Dice coefficient, Hausdorff distance and overall quality of the reproduced masks. Experimental results show that modified U-Net provides us with the best results for lumen and intima segmentation with Dice coefficient 0.985 and 0.98, and Hausdorff distance 0.194 and 0.101 mm, respectively.