Agreement-Based Semi-supervised Learning for Skull StrippingAugust 5th, 2010
FiledAugust 5th, 2010Supervised learning approaches have become increasingly popular and practi- cal in brain MRI segmentation [1,2,3]. These algorithms produce classifiers that utilize a large number of features by applying modern learning algorithms. How- ever, supervised learning often demands large amounts of training data with consistent manual labeling, which are difficult to obtain. Recent semi-supervised learning approaches [4,5,6,7] have provided new mechanisms to take advantage of the information in unlabeled data to train a better system.