Cardiac Image Segmentation for Contrast Agent Videodensitometry

Abstract

Abstract-Indicator dilution techniques are widely used in the intensive care unit and operating room for cardiac parameter measurements. However, the invasiveness of current techniques represents a limitation for their clinical use. The development of stable ultrasound contrast agents allows new applications of the indicator dilution method. Ultrasound contrast agent dilutions permit an echographic noninvasive measurement of cardiac output, ejection fraction, and blood volumes. The indicator dilution curves are measured by videodensitometry of specific regions of interest and processed for the cardiac parameter assessment. Therefore, the major indicator dilution imaging issue is the detection of proper contrast videodensitometry regions that maximize the signal-to-noise ratio of the measured indicator dilution curves. This paper presents an automatic contour detection algorithm for indicator dilution videodensitometry. The algorithm consists of a radial filter combined with an outlier correction. It maximizes the region of interest by excluding cardiac structures that act as interference to the videodensitometric analysis. It is fast, projection independent, and allows the simultaneous detection of multiple contours in real time. The system is compared to manual contour definition on both echographic and magnetic resonance images.

Quantification of Regional Left Ventricular Dyssynchrony by Magnetic Resonance Imaging

“…The filter is an improvement of that proposed in [36] by integration of multiscale analysis, so that no threshold definition is required.…”

“…In general, the endocardial boundary is characterized by a sharp gray-level drop at the transition between the blood pool and the myocardium. These transitions produce high frequencies in the frequency spectrum, which can be enhanced by a monodimensional high-pass filter along the rays originating from the seed point [36]. However, the edge detection in [36] needs the definition of a threshold, which remains a critical issue due to its dependency on the image characteristics.…”

“…These transitions produce high frequencies in the frequency spectrum, which can be enhanced by a monodimensional high-pass filter along the rays originating from the seed point [36]. However, the edge detection in [36] needs the definition of a threshold, which remains a critical issue due to its dependency on the image characteristics. Therefore, we propose the replacement of the radial filter in [36] by a more elaborated algorithm, which does not require the definition of a threshold.…”

“…In this study, we propose a development of the algorithm presented in [36]. This method, first applied in the context of the indicator dilution analysis for the definition of an optimal region of interest (ROI), showed sufficient accuracy to be extended to other applications.…”

“…The method in [36] detects the endocardium along a beam of rays originating in the ventricular center, referred to as a seed point. The seed point position is manually defined by the user.…”