NEURAL-NETWORK-BASED SEGMENTATION OF MULTIMODAL MEDICAL IMAGES - A COMPARATIVE AND PROSPECTIVE-STUDY

This paper presents a study investigating the potential of artificial neural networks for the classification of registered magnetic resonanc e and x-ray computer tomography images of the human brain. First, topo logical and learning parameters are established experimentally. Second , the learning and generalization properties of the neural networks ar e compared to those of a classical maximum likelihood classifier and t he superiority of the neural network approach is demonstrated when sma ll training sets are utilized. Third, the generalization properties of the neural networks are utilized to develop an adaptive learning sche me able to overcome inter-slice intensity variations typical of MR ima ges. This approach permits the segmentation of image volumes based on training sets selected on a single slice. Finally, the segmentation re sults obtained both with the artificial neural network and the maximum likelihood classifiers are compared to contours drawn manually.