Histopathology images are widely used by scientists as a data source in research involving artificial intelligence and deep learning. These images are constructed from scans of tissue specimens surgically extracted from patients, fixed onto glass slides, and stained with hematoxylin & eosin. It is important to note that during slide processing, pathologists often mark slides with colored pen to identify regions of interest. Although such slides may be clinically useful, when they are utilized in research, pen markings significantly interfere with the performance of computational models. More specifically, due to the prevalence of pen markings in datasets, models incorrectly learn to predict disease attributes based on the presence of these pen markings instead of human tissue characteristics. Through my prior research with the University of Pennsylvania’s AI in Biomedical Imaging Lab, I’ve discovered that there are no effective methods to eliminate pen markings from histopathology images. Consequently, researchers are forced to discard slides with such markings, markedly reducing dataset size. Given the already-limited nature of this data, especially for rare diseases, finding a solution is vital. Working with the lab at Penn, I am developing a computational image modification pipeline that accurately eliminates colored pen markings from histopathology images using color-altering techniques while preserving the original tissue data beneath the markings, thus maximizing model performance. My project’s success will allow researchers across the globe to stop excluding entire slides just because of pen markings, enabling the discovery of new medical findings that have the potential to save countless lives.