Optimal linear transformation for MRI feature extraction

TLDR

The method relates to discriminant analysis. The study develops and applies a feature extraction method for MRI that does not require explicit tissue parameter calculation. The method constructs a 3‑D feature space via a linear minimum‑mean‑square‑error transformation that clusters normal tissues at target positions and abnormalities elsewhere, then uses cluster histograms to define ROIs, estimate prototype vectors, and segment MRI scenes, demonstrated on simulations, an egg phantom, and a human brain. The proposed feature space outperforms tissue‑parameter‑weighted, principal component, and angle images in feature extraction and scene segmentation.

Abstract

This paper presents development and application of a feature extraction method for magnetic resonance imaging (MRI), without explicit calculation of tissue parameters. A three-dimensional (3-D) feature space representation of the data is generated in which normal tissues are clustered around prespecified target positions and abnormalities are clustered elsewhere. This is accomplished by a linear minimum mean square error transformation of categorical data to target positions. From the 3-D histogram (cluster plot) of the transformed data, clusters are identified and regions of interest (ROI's) for normal and abnormal tissues 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 tissue-parameter-weighted images, principal component images, and angle images, demonstrating its superiority for feature extraction and scene segmentation. Its relationship with discriminant analysis is discussed. The method and its performance are illustrated using a computer simulation and MRI images of an egg phantom and a human brain.