The Electrocardiogram (ECG) can be considered one of the most widely used devices in the investigation of heart activity and the identification of various cardiovascular diseases. In this regard, Deep Learning has been significantly developed since it allows automating processes performed manually by medical professionals. While there are many ECG datasets available, not a lot of them are accompanied by pixel-wise annotation for semantic segmentation purposes. This situation can be seen as another important problem of Deep Learning. The current research attempts to create a pipeline for creating synthetic images with segmentation masks from the signals stored in the PTB-XL dataset.