<title>MRI image segmentation using multiscale autoregressive model and 3D Markov random fields</title>

Abstract

ABSTRACT Texture segmentation applied to magnetic resonance imaging (MRI) is investigated using a multiscale autoregressive model(M-AR)1. Since M-AR models need large region for good parameter estimation, a mixture model using M-AR and constantgray level value is developed. Region uniformity is obtained using a 3D Markov random field (3D MRF). The segmentation isgiven by its maximum a posteriori (MAP) estimate. The segmentation is computed using iterated conditional modes (1CM).Two initial segmentation choices are studied: MLE segmentation with multiple resolution segmentation and human atlas.Human atlas initial segmentation proves to be closer to desired segmentation, even if the image from the atlas is not precise.KEYWORDS: human atlas, autoregressive model (AR), Markov random field (MRF), magnetic resonance imaging (MRI),segmentation, texture. 1. INTRODUCTION The objective of MRI segmentation is to help the specialist select relevant information from MRI sequences for diagnostic,treatment evaluation, etc. Most segmentation algorithms use a constant gray level model to represent the different structuralelements that can be observed in MRI. if we look closely at MRI images, we can see that some of its elements, like the femurin Figure 1, have correlated gray level variations. A more robust model than constant gray level should therefore be used torepresent these elements. This model must use the information provided by the gray level variation. A texture model wouldtherefore be appropriate. Texture is defined as a structure composed of a large number of more or less ordered similar