MRI image segmentation via neural network fusion

We employ an artificial neural network to fuse a tripler of "multi-spectral" brain images front a magnetic resonance imaging system into a segmented image. The pixel values for the same pixel location from each of T1, T2 and PD images of the same slice of a given brain scan are input to a neural network for training. The three output components each rake high or low values to form codewords for different grayscale classes. Eighty pixel locations from each class are sampled as triplets (T1, T2, PD) and used for backpropagation training. Then the network maps each novel triplet into art output codeword that represents one of the 6 class grayscales and that grayscale is put into that pixel location in the output image. Other researchers have mapped triplets of representative values, e.g., of medians over small blacks, bur this has the effect of oversmoothing and blurring the segmented regions. Our method appears to be more practical.