Development of automatic recognition software of left ventricle by time series processing echocardiograms and application to disease heart

Many techniques to recognize the shape of the left ventricular cavity have developed to improve the cardiac evaluation, and some of the techniques are applied to a clinical diagnosis now. However, most of these techniques require manually initial settings. Therefore, we have developed an automatically recognition algorithm without initial settings. This algorithm automatically detects the contour of left ventricular cavity by considering all of the combinations between motion vectors of the ventricular wall to extract intersection points of vectors. We have applied this algorithm to 10 normal hearts and 60 diseased hearts. As the result, the proposed method correctly recognized 85% of left ventricular cavities throughout all of the recorded frames, which was judged by professional sonographers. Furthermore, this software was applied to 40 normal hearts and 113 disease hearts where automatic calculation of ejection fraction was agreed with the results of conventional method by a sonographer. Finally this software was applied to real-time visualization of the cavity through analog video signal.