In recent years non-invasive medical diagnostic techniques have been used widely in medical investigations. Among the various imaging modalities available, Magnetic Resonance Imaging is very attractive as it produces multi-slice images where the contrast between various types of body tissues such as muscle, ligaments and fat is well defined. The aim of this paper is to describe the implementation of an unsupervised image analysis algorithm able to identify the body fat tissues from a sequence of MR images encoded in DICOM format. The developed algorithm consists of three main steps. The first step pre-processes the MR images in order to reduce the level of noise. The second step extracts the image areas representing fat tissues by using an unsupervised clustering algorithm. Finally, image refinements are applied to reclassify the pixels adjacent to the initial fat estimate and to eliminate outliers. The experimental data indicates that the proposed implementation returns accurate results and furthermore is robust to noise and to greyscale in-homogeneity.