Echocardiography is a non-invasive method for monitoring ventricular function and evaluating myocardial hypertrophy -in patients with heart disease. Computer-assisted image analysis methods are designed to facilitate and standardize the extraction and quantification of anatomical and physiological parameters that are critical for diagnostic purposes. We have previously developed image processing techniques for the analysis of one-dimensional or M-mode echocardiography, the mode widely accepted as the reference method for the evaluation of myocardial wall thickness. Our techniques address two types of problems in particular: (i) the automated extraction of myocardial borders, and (ii) the restoration of sequences of several consecutive heart beats by performing an average with respect to a normalized time-scale. More recently, we have extended some of these techniques for the detection of endocardial borders in two-dimensional echocardiograms. An algorithm that we propose is designed to detect boundaries that can be parameterized in polar coordinates. Given the specification of a center point, the image data is recoded in polar coordinates and the border characteristics are enhanced by matched filtering. The final processing step is a border tracking algorithm that uses dynamic programming optimization. our preliminary results are extremely promising and suggest that the method is robust.
