To evaluate truthing (also known as label fusion) methods in medical image segmentation, synthetic segmentation contours can be useful especially when the reference standard is established by combining multiple segmentation results, such as those produced by multiple experts. This is because ground-truth segmentation is often unavailable in real medical images but is predefined in synthetic data. For this purpose, we developed the Restorable Segmentation Synthesis (RSS) tool. The RSS tool generates segmentation contours by modifying the Fourier descriptors of a truth contour, which, for realism, can be the contour of an anatomical structure extracted from a real medical image. The tool allows for the creation of contours with various segmentation errors relative to the ground truth. A favorable feature of our segmentation contour synthesis tool for evaluating truthing methods is that the average of a large number of synthetic contours asymptotically converge to the truth contour. This is important because such a dataset can help benchmark and compare the truthing methods. Our RSS tool is developed to have this restorability property, which we validated here through simulation studies. We further show that simulating contours is a promising approach for truthing method analysis and data augmentation for segmentation tasks. The RSS tool with a GUI is available: https://github.com/DIDSR/RSS-tool