METHOD FOR AUTOMATIC SEGMENTATION OF BRAIN TUMORS MERGING FULL CONVOLUTION NEURAL NETWORKS WITH CONDITIONAL RANDOM FIELDS

A method for automatic segmentation of brain tumors merging full convolution neural networks with conditional random fields. The present application intends to address the issue that presently the technology of deep learning is unable to ensure the continuity of the segmentation result in shape and in space when segmenting brain tumors. For this purpose, the present application includes the following steps: step 1, processing a magnetic resonance image comprising brain tumors by utilizing a method for non-uniformity bias correction and brightness regularization, to generate a second magnetic resonance image; step 2, performing brain tumor segmentation for said second magnetic resonance image by utilizing a neural network merging a full convolutional neural network with a conditional random field, and outputting a result of brain tumor segmentation. The present method may execute brain tumor segmentation end-to-end and slice by slice during testing, which has relatively high computation efficiency.