Author(s)

Y. Tada & A. Shibata

Abstract

It is necessary to recognize major tissue in images of MRI of brain for an au- tomatic diagnosis. In this paper, we outline a scheme for automatic extraction of main tissue by classification of pixels based on the information of density and position in the image. The information of position is given by the membership function so that it may have a large value for a pixel where the probability that the tissue exists is high. We propose two techniques for classification; one is to use a back propagation network which has studied the training data and the other is to minimize an energy function by a Hopfield network. We use pairs of Tl-weighted and T2-weighted images by MRI as the data for the classification. The result of classification is good enough.

Keywords