MACHINE INTELLIGENCE FOR BRAIN SEGMENTATION: A TOOL TO IDENTIFY BRAIN ILLNESSES THROUGH SEGMENTATION
Machine Learning is becoming a prominent force in the medical field. We created Machine Intelligence for Brain Segmentation (MIBS), a tool that segments brain MRIs into different colors that signify enhancing tumors, non-enhancing tumors, and edema. The dataset used was of 624 MRIs from the Medical Segmentation Decathlon. The model was trained with the U-Net algorithm, a Convolutional Neural Network made for Biomedical Image Segmentation, and resulted in an average accuracy of ~99% across the different classes and ~0.75 an average F1- score across the different classes.
READ FULL TEXT