The main goal of the project was to develop an algorithm for detection of text and labels added artificially to chest X-ray images. By using this algorithm, we can improve the accuracy of other algorithms that are used for giving medical diagnosis based on chest X-ray images. We chose using a deep learning approach to solve this object detection problem. This approach was chosen due to its capability for extracting automatically high-quality and complex features straight from the input data. Deep learning networks require an input dataset on which they will be trained and tested. Since our problem did not have a proper dataset with the relevant tags, we had to create one ourselves. Afterwards, we built a Fast R-CNN based model. We reached an average precision of 99.85% on the training set and 99.86% on the test set, and we managed to achieve good results even for images out of our database, for which we got 97% true positive detection.