Partial Volume Effect Detection in Mri Segmentation Based on Approximate Decision Reducts

Segmentation of Magnetic Resonance Imaging (MRI) is a process of assigning tissue class labels to voxels. One of the main sources of segmentation error is the partial volume effect (PVE) which occurs most often with low resolution images – with large voxels, the probability of a voxel containing multiple tissue classes increases. We propose a multistage algorithm for segmenting MRI images with a mid-stage of recognizing the PVE voxels. The information about PVE regions added to other voxels features extracted from the image can increase the overall accuracy of the segmentation. In our methods we have utilize a classification approach based on approximate decision reducts derived from the data mining paradigm of the theory of rough sets. An approximate reduct is an irreducible subset of features, which enables to classify decision concepts with a satisfactory degree of accuracy in the training data. The ensembles of best found reducts trained for appropriate approximation degrees are applied to detection of the PVE and performing the segmentation.