Study of Brain Region Segmentation Using Convolutional Neural Network

Magnetic Resonance Imaging (MRI) is used in medical imaging for detection of tumours and visualize brain tissues. This is done manually by expert radiologist and this takes good amount of time. The traditional method of MRI evaluation of tumour depends greatly on qualitative features, like density of tumour, growth pattern etc. Brain Region Segmentation is important in neuroimaging application, for example, alignment of images, surface reconstruction etc. The previous methods depends upon the qualitative features and is very sensitive to errors. Noise and errors need to be reduced and efficiently delineated, very less work is done in automatic tumour detection using deep learning methods and there is lot of areas which can be explored. The deep learning method is very much different from the machine learning method. The machine learning method uses algorithms to input data, learn from given data, and make decision based on the experience or learning whereas the deep learning can learn and make decisions on its own. Deep learning has a capability of learning from data that is unstructured or unlabeled. In deep learning, the algorithms try to learn using method of feature extraction which is very different and makes the model fully automatic, here we don’t require any handcrafted feature. In traditional method we need to develop feature extractor for different problem, so we use deep learning which reduces effort of developing different feature extractor for different problem. In one of method of 2-