Research On Image Segmentation Technology Of Left Ventricular Magnetic Resonance Sequence Based On Learning-driven

Cardiovascular disease,which attracts much attention from doctors,has been one of the main threats to human health.As the main means for doctors to diagnose cardiovascular diseases,medical imaging technology has been applied in the diagnosis,treatment guidance and cardiac function assessment before and after operation.In order to obtain good cardiac function parameters,the accurate segmentation of each cardiac chamber should be completed on the image.In the heart MR image,the gray value of left ventricular myocardium is similar to that of trabeculae,mastoid muscle and surrounding muscle tissue,leading to the existence of weak edge around the myocardium,which makes it difficult for many algorithms to accurately segment the left ventricle.In order to solve the above problems,the segmentation of left ventricle will be studied in this paper.The specific research contents are as follows:(1)Motion target detection method was used to track the position of the left ventricle,the Region of Interest(ROI)containing the left ventricle was clipped,and subsequent studies were conducted on the basis of ROI.(2)In the ROI image,when the left ventricle is segmented by the Distance regular level set(DRLSE)algorithm,there are problems of image gradient information dependence and initial contour sensitivity.Based on DRLSE,this paper proposes a left ventricular segmentation algorithm with weak edge information.The algorithm USES the fitting method to calculate the new local terms based on the coefficient of variation segmentation model(PSM),and reduces the sensitivity of DRLSE to the initial contour.According to the shape priori of the inner and outer membrane,the velocity function that drives the evolution of the curve is introduced to overcome the leakage problem of the DRLSE algorithm at the weak boundary of the outer membrane of the left heart.(3)The level set method combined with neural network is studied,and the algorithm in CRBM drive(2)is used.First,CRBM training is completed by combining new data sets.Secondly,using the trained CRBM to predict the next evolution shape of the curve,form shape constraints and add them into the algorithm(2).Finally,through the analysis of left ventricular myocardium thickness,another velocity function of curve evolution was proposed to replace the velocity function of(2).To verify the accuracy of the proposed algorithm segmentation,this paper based on the image of hospital in Toronto sick children to provide public database,the algorithm presented in this paper the ROI image segmentation,the experiment segmentation results are qualitative assessment,the results show that the presented segmentation algorithm,can reduce the DRLSE is sensitive to the initial contour,improve contains precision segmentation of weak edge information of the left ventricle.