OPTIMAL LINEAR TRANSFORMATION FOR MRI FEATURE-EXTRACTION

This paper presents development and application of a feature extractio n method for magnetic resonance imaging (MRI), without explicit calcul ation of tissue parameters. A three-dimensional (3-D) feature space re presentation of the data is generated in which normal tissues are clus tered around prespecified target positions and abnormalities are clust ered elsewhere. This is accomplished by a linear minimum mean square e rror transformation of categorical data to target positions. From the 3-D histogram (cluster plot) of the transformed data, clusters are ide ntified and regions of interest (ROI's) for normal and abnormal tissue s are defined. These ROI's are used to estimate signature (prototype) vectors for each tissue type which in turn are used to segment the MRI scene. The proposed feature space is compared to those generated by t issue-parameter-weighted images, principal component images, and angle images, demonstrating its superiority for feature extraction and scen e segmentation. Its relationship with discriminant analysis is discuss ed. The method and its performance are illustrated using a computer si mulation and MRI images of an egg phantom and a human brain.