Segmentation Modalities of Gliomas in MRI Scans
By: Alex Sanchez, Cevat Güleç
Department: Computer Science
Faculty Advisor: Dr. Ilmi Yoon
Through the combination of Google Colaboratory and machine learning, we investigate whether we can train an UNet model and predict a whole brain tumor through training, testing and validation of batches of patient glioma nifti images. With a set of objectives, we set to pre-process said images, build a UNet model, train said model and be able to predict whether a test image has glioma as well as identifying any possible tumor subregions. In addition, we will look the learning curve, and whether improvements in hyperparameters will increase the possibilities of identifying glioma and subregions with the help of segmentation mask.