Full Myocardium Segmentation Of The Echocardiographic Images And Myocardial Infarction Diagnosis

In this paper we propose a full-automatic method of segmenting the epicardiumand endocardium of echocardiography images, we also propose a diagnostic methodof myocardial infarction. The data used in myocardial infarction diagnosis isacquired from the time-intensity curve of the whole myocardial perfusion process.And each data point corresponds to one echocardiographic image, because the datais the mean gray value of the myocardium region of corresponding image.The segmentation of the endocardium is mainly based on mathematicalmorphology. And while segmenting the epicardium Chan-Vese active contour modelis used. Because the left ventricle cavity is only adjacent to the myocardium, and thecavity region is similar in gray value, so the simple mathematical morphologymethod can have good performace. But the epicardial region is adjacent differenttissues and is high-noise region, so the mathematical is not an ideal method. But themethod of Chan-Vese active contour model which is based on region and is notsensitive to noise, so we use this method to acquire the initial contour. And we usethe fitted circle to connnect the overflowing region.While diagnose the myocardial infarction HHT method is used. First weseparate the myocardium region into six section, and then decompose the curve ofdifferent region with EMD in order to get the intrinsic mode functions, then useHilbert to get the instantaneous parameters. Finally we get the marginal spectrum.Experiment with normal myocardial regions and myocardial infarction regions, wehave a conclusion: while the energy of marginal spectrum is concentrated on1-2Hzand6-8Hz, then this region is normal myocardial region, and when the energy isvery small or is only concentrated on1-2Hz, then this region is myocardialinfarction region. And frequency range1-2Hz and6-8Hz represent the respiratoryfrequency and heart rate respectively.The reason we use HHT is that there are many signals all around the humanbody, like heart rate, respiratory frequency and brain wave. So we guess thetime-intensity curve will be inflluenced by human body signals. And the key stepEMD of HHT method can decompose signal to signals of different frequency, so weuse this method to decompose the signal and finally we propose a diagnosis methodof myocardial infarction according to the marginal spectrum analysis.