Probabilistic Multi-Modality Image Segmentation with Partial Voluming
Maja Pokrić, Paul A. Bromiley, Thacker Na, Aaron S. Jackson · 2002
This paper addresses the creation of tissue models for multi-modal medical images. Our novel multi-dimensional model represents image intensities using contributions to the signal from pure tissues and partial mixtures. Expectation Maximisation (EM) optimisation technique is used to iteratively adjust model parameters, while Bayes theory is applied to generate tissue voluem fraction maps for each segmented tissue. The technique is useful for the fusion of information in images as well as for graphical rendering.