Automated detection of the left ventricular region in gated nuclear cardiac imaging.

An approach to automated outlining the left ventricular contour and its bounded area in gated isotopic ventriculography is proposed. Its purpose is to determine the ejection fraction (EF), an important parameter for measuring cardiac function. The method uses a modified version of the fuzzy C-means (MFCM) algorithm and a labeling technique. The MFCM algorithm is applied to the end diastolic (ED) frame and then the (FCM) is applied to the remaining images in a "box" of interest. The MFCM generates a number of fuzzy clusters. Each cluster is a substructure of the heart (left ventricle,...). A cluster validity index to estimate the optimum clusters number present in image data point is used. This index takes account of the homogeneity in each cluster and is connected to the geometrical property of data set. The labeling is only performed to achieve the detection process in the ED frame. Since the left ventricle (LV) cluster has the greatest area of the cardiac images sequence in ED phase, a framing operation is performed to obtain, automatically, the "box" enclosing the LV cluster. THe EF assessed in 50 patients by the proposed method and a semi-automatic one, routinely used, are presented. A good correlation between the two methods EF values is obtained (R = 0.93). The LV contour found has been judged very satisfactory by a team of trained clinicians.

Delineation and quantitation of brain lesions by fuzzy clustering in positron emission tomography.

In this study, we investigate the application of the fuzzy clustering to the anatomical localization and quantitation of brain lesions in Positron Emission Tomography (PET) images. The method is based on the Fuzzy C-Means (FCM) algorithm. The algorithm segments the PET image data points into a given number of clusters. Each cluster is an homogeneous region of the brain (e.g. tumor). A feature vector is assigned to a cluster which has the highest membership degree. Having the label affected by the FCM algorithm to a cluster, one may easily compute the corresponding spatial localization, area and perimeter. Studies concerning the evolution of a tumor after different treatments in two patients are presented.