Unifying Statistical Classification and Geodesic Active Regions for Segmentation of Cardiac MRI

Abstract

This paper presents a segmentation method that extends geodesic active region methods by the incorporation of a statistical classifier trained using feature selection. The classifier provides class probability maps based on class representative local features, and the geodesic active region formulation enables the partitioning of the image according to the region information. We demonstrate automatic segmentation results of the myocardium in cardiac late gadoliniumenhanced magnetic resonance imaging (CE-MRI) data using coupled level set curve evolutions, in which the classifier is incorporated both from a region term and from a shape term from particle filtering. The results show potential for clinical studies of scar tissue in late CE-MRI data.

Segmentation of the left ventricle in cardiac cine MRI using a shape-constrained snake model

“…The prior information may be the statistical shape from a training set [10,26,27], be anatomical information such as an ellipse [28][29][30], or be intensity statistics [31,32]. Paragios [26] introduced an intensity consistency energy constraint into the variational level set approach [33].…”

“…Paragios [26] introduced an intensity consistency energy constraint into the variational level set approach [33]. Folkesson et al [27] presented a segmentation method that extends the geodesic active region method by the incorporation of a statistical classifier trained using feature selection. Ben Ayed et al [11] proposed to get curve evolution equations by minimizing two functionals each containing an original overlap prior constraint between the intensity distributions of the cavity (a) The segmentation results of the LV without shape based constraints (b) The segmentation results of the LV with shape based constraints Fig.…”

“…and myocardium. Although excellent results have been achieved in [26,27,11] where the LV shape is learned from an annotated training data set, the segmentation performance depends heavily on the size and richness of images in the training set.…”

A Fast Convex Optimization Approach to Segmenting 3D Scar Tissue from Delayed-Enhancement Cardiac MR Images

“…Recently, several approaches [2,3] were proposed to segment the 3D scar tissue, which employ additional information from other images to constrain the search for scar tissue within the myocardium geometrically. Additional myocardial segmentations [4], or registrations [5] to other images is time consuming and a potential source of error.…”

Guest Editorial Introduction to the Special Section on Computer Vision for Intravascular and Intracardiac Imaging

“…They incorporated a supervised learning in terms of a statistical classifier into a geodesic active region framework in order to particularly deal with irregular appearance due to scar tissue in the myocardium. A shape prior obtained from a shape particle filtering is also combined into the final segmentation for the epi-and endocardium with an additional coupling term for the two contours [13].…”