A Novel Distance Transform for Brain Extraction from T1-W Magnetic Resonance Images (MRI) of Human Head

Extracting the human brain tissues from MRI is a vital pre-processing step required for many automated medical image-processing pipelines. We propose two methods NDTM and NDTME based on the Novel Distance Transform (NDT) technique to identify the brain boundary in a 2D MRI and separate it. These methods utilize the largest connected component analysis and mathematical morphology. Experiments are done by applying these two methods on thirtyeight sets of MRI head scans (3500 images) collected from the publicly available Internet Brain Segmentation Repository (IBSR) image database. Achievement of these methods is assessed by estimating the Jaccard and Dice similarity indices. It was found that our method NDTME produced higher precision values than BSE, BET, and a few other algorithms. The suggested algorithms are simple, noise tolerant, and have lower computational complexity than the existing methods.