Fuzzy Clustering for Parametric Map Construction in Myocardial Perfusion Magnetic Resonance Images

Abstract

Ischemic heart disease is due to an imbalance of blood flow demand and supply. In clinical settings, nuclear imaging modalities, Single Photon Emission Computed Tomography (SPECT) or Positron Emission Tomogaphy (PET), are used to detect blood flow (perfusion) defects in the myocardium. However, Magnetic Resonance Imaging (MRI) presents the advantage of a better spatial resolution than nuclear methods and a sufficient temporal resolution to detect small perfusion defects that may not be detected by PET or SPECT. Since perfusion analysis can only be derived from time intensity diagrams over a series of images, computer-assisted image evaluation is required. A parametric map that summarizes the temporal behavior of the pixels within the heart image is necessary. In nuclear medicine, the parametric map construction is based on a model of the contrast agent used to image the myocardial perfusion over time. In MRI, the definition of such a model is still a challenging issue. This work shows that an unsupervised fuzzy clustering technique may be a good alternative to build a perfusion parametric map. Being unsupervised, this technique does not require a model, and the notions of fuzzy clusters and membership degree are adequate to show the smooth transition between healthy and pathological tissues. However, a careful use of similarity measures for the clustering process is required.