Research On Echocardiographic Spot Tracking Algorithm

With the increase of the proportion of aging population in China,the incidence of cardiovascular disease has also increased year by year,prevention and early treatment is very important for the control of cardiovascular diseases.Ultrasound diagnosis has become one of the main methods of medical imaging technology due to its real-time imaging,no harm,no pain and low cost.Echocardiographic speckle tracking is based on the two-dimensional image,the spot area of the left ventricular wall is identified and extracted,and then the motion is tracked in each frame of the image sequence,and finally the spot motion track is drawn through the spot.The displacement status is an assessment of cardiac health status.In this paper,the left ventricle of the heart was first segmented,and the purpose was to automatically identify and extract the spots of the myocardial wall of the heart based on the segmentation results.The segmentation algorithm consists of three parts.Firstly,the region of different echo intensity of the heart was divided into independent parts by the three-phase level set method.Then,the ventricular wall region was extracted by the binary processing method,the noise and the myocardial wall area were removed and connected respectively.Finally,the left ventricle contour was fitted by curve fitting method and segmented into a smooth closed segmentation curve.The algorithm segmentation results were compared with the doctor's manual segmentation results,and the left ventricle was qualitatively segmented.The results were evaluated using three image segmentation evaluation methods: RDD,ROD,and Dice parameters.The RDD value is 5.1%,ROD and Dice parameter values both are close to 90%,providing quantitative results of segmentation of the left ventricle.The segmentation algorithm has a good segmentation effect on the left ventricular contour,and it is not sensitive to intracardiac noise;the processing under the binary image removes most of the left ventricular cavity noise while retaining the information of the ventricular wall region.In the curve fitting process,the sampling of the ventricular wall is fitted to complete the smoothing of the final segmentation effect,which reduces the difference of the results under different images,thereby improving the stability of the left ventricular segmentation.The left ventricular wall spots were identified based on the segmentation results and then tracked.Firstly,the influence of the two parameters of search window size and image block size was analyzed,and then the block matching method was used to smooth the constraints,in this paper,the block matching method was constrained by the idea that Horn-Schunck assumes that the optical flow changes smoothly on the image.Then the pyramid block matching method was introduced,and the pyramid was constructed from the bottom layer to the upper layer by the method of accumulating adjacent pixels.The pyramid block at the point to be matched in the search window was compared from top to bottom to measure the similarity.Finally,analyze and compare the performance of the full search method and the logarithmic search method.The accuracy of the experiment was obtained by comparing the result with the known parameters of the image shift.Under the premise that the image block size is fixed,if the size of search window is not less than the known image shift amount,the accuracy is greater than 90%,if the size of search window is less than the known image shift amount,the accuracy is 0.The timeconsuming increases with the increase of the search window;under the premise that the search window size is fixed,the accuracy rate increases first and then decreases as the image block increases.It reaches 93.1% at the maximum value,and the time-consuming increases with the increase of the size of image block.From the result analysis,the search window must be larger than the actual displacement value of the image to ensure the tracking accuracy.When the image block size is at an intermediate value,the accuracy of the tracking results is the best,and the time consumption increases in multiples as the size of the search window or the image block increases.The block matching method under optical flow constraint is improved by 5.3% compared with the traditional block matching method.However,its time consumption is increased by nearly 4 times.The optical flow constraint can improve the accuracy of block matching,but at the same time increase the redundancy of the algorithm.The pyramid block matching method optimizes the algorithm speed by nearly 3 times while maintaining the accuracy of the traditional block matching.The logarithmic search method maintains the accuracy of the full search method in the case of a reasonable setting of the search window size,and the algorithm is optimized by 2 to 4 times.At the end of the article,the research method is used for clinical application.The result of the five volunteer heart image sequences shows that the hearts have regular movement and the displacements are in the range of 4-8 mm.The results of the algorithm in this paper are within the normal range of motion parameters in clinical application.