Development of recognition software of heart to find the standard cross section on echocardiography

We have developed an algorithm to find standard cross sections (the long-axis view and the short-axis view) of the heart from successive echograms. We first divided an echogram into small spatial regions and detected the typical motion of the mitral valve by analyzing the brightness variation and correlation coefficient among the regions. We have obtained 95% accuracy in the position of the valve through time series echogram of 25 normal volunteers. The recognized valve was visualized as a mark on the video stream. Furthermore, combining this technique with an optical flow method, we elucidated the region velocity of the wall motion of the left ventricle after centering the valve on echogram. By analyzing symmetry among region velocity, we have confirmed to distinguish between the long- and the short-axis view of heart. This algorithm is applicable to instruction software to find standard cross section of the heart as an assistant of echocardiography. We are going to apply to more subjects who have heart disease and to contribute automatic diagnosis in the future.