Neural Network Based Medical Image Processing.doc

 PAGE \* MERGEFORMAT 6 Neural Network Based Medical Image Processing [Abstract] described a neural network-based medical image compression method. [Keywords:] neural network; image processing; BP algorithm 70 years beginning from the 20th century, with the invention of CT, the medical image processing in the medical field occupies an increasingly important position. Through the images provide information on a variety of related diseases, clinicians can more directly examine lesions, significantly improved diagnostic accuracy. Medical image data are often very large amount of data, which gives the image distance transmission and storage of great inconvenience, so data compression is necessary, we will be a huge amount of image information compression, only the transmission or store compressed information, when the need to extract the image and then when certain rules of the images resumed. Neural network parallel processing and adaptive ability to learn to become a good medical image data compression tool. Neural network composed of image data compression system as follows: Input layer middle layer output layer This is a three-layer network, input layer and output layer have the same number of nodes, the middle tier of nodes than the input layer or output layer of nodes less. The network training process the network input layer and output layer use the same signal as a learning sample of images, due to the input layer and output layer use the same image signal, so the network can reproduce the input layer output layer signals. As the middle layer unit number considerably less than the input layer and output layer, the middle layer can be a valid indication of the input image signal transmission and storage. Input layer to the middle tier of the transformation can be seen as compressed data encoding process, the middle layer to output layer is a decoding process of transformation. Assumed to be encoded input image format for the M * N pixels points, to