When creating models with machine learning, overfitting is a common problem. That issue is particularly relevant when the models overfit to confounding factors which can be related to the subject characteristics or to imaging characteristics unrelated to the subject. In order to minimize confounding variable related to image acquisition properties, the author proposes a library and pipeline to normalize imaging data -- with development focus on biomedical imaging --, and a pedagogic tool to study how the several preprocessing steps affect the final normalized result. The tool was developed in Python Dash, as a web server, allowing the interaction with the tool through an user friendly web page.