SEGMENTATION OF MAGNETIC RESONANCE MICROIMAGES OF TRABECULAR BONE: CLASSIFIERS AND MARKOV RANDOM FIELD MODEL
I.Strolka Istrolka, A.Accardo Aaccardo, D.Dreossi Ddreossi, F.Vittur Fvittur, R.Toffanin Rtoffanin, I.Frollo Ifrollo · Measurement Science Review · 2003
Quantitative assessment of trabecular bone structure based on magnetic resonance microimages requires a segmentation step, which is difficult to perform because of low signal-to-noise ratio and spatial signal inhomogeneities in these images. In this paper, we present the design of voxel classifiers based on statistical mixture models and classifiers using the feed-forward artificial neural networks (ANN). In both cases a Markov random field (MRF) prior model is used to enhance the reliability of the segmentation process.