BRAIN TUMOR AUTOMATIC SEGMENTATION METHOD BY MEANS OF FUSION OF FULL CONVOLUTIONAL NEURAL NETWORK AND CONDITIONAL RANDOM FIELD

摘要

A brain tumor automatic segmentation method by means of fusion of a full convolutional neural network and a conditional random field. The method is to resolve the problem of failure to ensure the continuity of a segmentation result on the appearance and the space when the brain tumor segmentation is performed by using the existing depth learning technology. The method comprises the following steps: step 1, processing a magnetic resonance image comprising a brain tumor image by using a non-uniform offset correction and luminance regularization method, so as to generate a second magnetic resonance image; and step 2, performing brain tumor segmentation on the second magnetic resonance image by using a neural network fusing a full convolutional neural network and a conditional random field, and outputting the brain tumor segmentation results. By means of the method, brain tumor segmentation can be performed in an end-to-end and slice-to-slice manner, and accordingly the method has a higher operational efficiency.