The Algorithm Research Of The 3D Cardiac Twisting Motion For Ultrasonic And MR Image

Heart disease, one of the major causes of death, plays a considerable role in endangering human health at present. Cardiac twisting motion reflects the systolic function of left ventricle. Studies have shown that the extent and form of the cardiac twisting would be affected by some heart diseases. Thus, the research of the reverse movement of myocardial, which is on behalf of the functional status of heart during the cardiac cycle, is extremely important for determine whether it is normal for the heart. The analysis of the myocardial twisting angle provides a valuable basis for the clinical diagnosis of heart disease.The detection method of cardiac three-dimensional twisting motion, which is based on image processing, is proposed in this paper, including analysis of the DICOM image data, segmentation and contour tracking of left ventricle, extraction and matching the feature points on the left ventricular contour, three-dimensional reconstruction of the left ventricular cavity and analysis the projection of motion vector. Experiments have proved the feasibility of the detection system to detect the twisting angle of cardiac in three-dimensional space, breaking through the limitations of two-dimensional spatial analysis. Meanwhile, the algorithms designed for all the modules in the system have some innovation and effectiveness, meeting the research needs very well.This study mainly includes the following aspects:①The author have extracted the image data and three-dimensional information from the DICOM format medical files according to the DICOM 3.0 standard and then parsed the sequential frames of the sagittal,coronary, transection plane and converted them into BMP format images.②After comparing the detection results of traditional gradient edge detection operators, the author brought in an improved mathematical morphology edge detection method and constructed a good edge detection operator, solving the coordination problem between the edge detection accuracy and the anti-noise performance through the combination of different morphological operators and combining the characteristics of multi-structuring elements and multi-scale. Experiments have shown that the algorithm is able to filter out the noise and maintain the consistency and accuracy of the left ventricle's edge at the same time, which have also provided an accurate edge information map for future study. ③The improved fuzzy C-means clustering method, GVF Snake algorithm and Water Balloons Snake model were used in this paper to segment the left ventricular effectively and extract the outline of the left ventricular wall accurately. Fuzzy C-means clustering method contains the fuzzy theory, including the information of gray-scale, gradient and location. This method not only has prior information but also can distinguish the complex border easily. Based on the active contour model, the Snake segmentation algorithm positions the initial contour by the bound of area firstly, and then auto-closes the true outline of the objective through energy restriction, finally, the outline which is coherence and closer to the true contour is extracted. This method can overcome the difficulties of segmentation owing to the vague left ventricular region and the mixing of edge and shadow; furthermore, it also can repair the breakpoint of contour caused by noise and make the outline smooth and continuous. This algorithm is also able to trace the contour of consecutive frames at one time and is suitable for medical image segmentation with the significance for research.④The matching of feature points for sequence contour sections in three-dimensional space was completed, which combined the geometry feature and the dissection feature. Based on the greatest similarity of curvature and distance, this article posed the feature points matching algorithm which can find the correspondence of feature point in space more accurately. In this paper, the matching of feature points of consecutive frames slice in a cardiac cycle was completed by the BP neural network algorithm and this algorithm also can find the corresponding points of other points on the contour automatically with the help of a few matched feature points, greatly improving the computational efficiency. Finally, the author have calculated transformation parameters of space through the control point equation, obtained the twisting angle and displacement vector of objective's outline on adjacent sections of time series, tracking the movement of the left ventricular thereby.⑤In the platform of Open GL, the author achieved the three-dimensional reconstruction and display of the left ventricular cavity with three-dimensional surface reconstruction algorithm based on the splicing of contour. And then preliminary studied of the reconstruction of space displacement vector according to the relationship of orthogonal projection in the Cartesian coordinate system, and explored the three-dimensional motion vector calculation method.Experiments have proved that the cardiac reverse twisting motion algorithm system in this paper is feasible and the theory and algorithms involved have a certain reference value. We are looking forward to improve the algorithm and use them in the clinical diagnosis in the future.