Using neural networks in color kinesis image process to automate diagnosis of cardiac disease

Ultrasound has become an important tool for diagnosing cardiac disease, used to produce echocardiographic images. Physicians can visually diagnose heart disease by observing the movements of the heart walls on the echocardiogram. But visual diagnosis requires a trained physician, and is a matter of subjective interpretation, so that there is a possibility of differences of opinion between physicians. Color kinesis is a new technique for displaying echocardiograms. In color kinesis, the position of the ventricular endocardium is measured from the strength of ultrasonic reflections at frequent intervals, and its position is shown in different colors for each time slice. The echocardiogram generated by color kinesis, or the color kinesis image, can show in a single frame information on the movement of the ventricle walls from the beginning of a systole to the end. This is an effective way of diagnosing any abnormalities in the movement of the heart walls. In the investigation reported here we extracted the characteristics of the movements of the ventricular endocardium based on two color kinesis frames before and after administering a dobutamine load (to increase systolic force), and attempted to identify abnormalities based on the endocardium's movement characteristics before and after dosing. Identification was performed by a neural network.