Automatic left ventricular feature extraction and visualisation from echocardiographic images

A fully automatic system for extraction and visualisation of the cardiac Left Ventricular (LV) features such as ventricular volume, ejection fraction, thickness of the myocardium and wall motion from two-dimensional echocardiographic images is presented. The system has the ability of processing both the Short Axis (SA) and Long Axis (LA) standard echo views. In the proposed system fuzzy logic is used to automatically define a central point inside the LV cavity using knowledge of the standard echo views and the heart morphology. A robust multiresolution edge detection technique based on wavelet transform is then used in conjunction with a radial search algorithm, to define a candidate single boundary point on each radial intensity profile emanating from the predefined central point. Both spatial and temporal median post filtering and Cubic B-spline approximation methods are employed to refine the estimated boundary points and to construct the final closed sequence of LV boundaries. These boundaries together with some appropriate approximating formulae are used to calculate and visualise the above mentioned LV features. Experiments and results with real SA and LA echo images (supplied by Cardiology Department of the Western Infirmary, Glasgow) are presented.