Research Of Heart Images Segmentation Based On Graph Cuts And Active Shape Model

The present domestic and international trends in the incidence of heart disease increases gradually, heart health issues that cannot be ignored The ultrasonic diagnostic has characteristics such as noninvasive, non-injury, non-ionization radiation to human body, cheap, real time and so on advantages, which making it one essential diagnostic technology in modern clinical medicine. However, due to the intrinsic characteristics of ultrasonic, the ultrasound image contains a large number of speckle noise and other noises; this increases the difficulties to image processing, diagnosing quantitative analysis and so on, while the image segmentation is the first step of all medical image processing.Based on this, this paper mainly studied the use of graph cuts algorithm and the active shape model on cardiac MRI images and ultrasound images segmentation, and did analysis on experimental results.Graphs cut algorithm based on graph theory and has a good mathematical foundation. First it builds energy function and s-t network, utilizes the maximum flow/minimum cut theorem, minimize energy function to solve the segmentation problem. In this paper, through theoretical research and experimental analysis I found that the algorithm combines with the subsequent image processing did a good job on magnetic resonance heart images, but not for ultrasonic images very much. To make the experiment more rigorous, this paper added image preprocessing that suit for ultrasonic images. The segmentation results improved, but not ideal.Active shape model mainly applied on the dynamic recognition, face recognition, the shape of the hand segmentation etc. rarely used on medical ultrasonic image. ASM mainly added into the prior knowledge through training, got the initial model, match again, an iterative update parameters to find the segmented regions. This algorithm can be used in clinical ultrasound images directly without the preprocessing steps, reserve original image information to the greatest extent. In this paper, I studied the basic theory of active shape model and then did lots of experiments and improvements to make the algorithm suitable for heart ultrasound images'segmentation. ASM got better segmentation results than graph cuts for cardiac ultrasound image processing.After collecting of several consecutive cardiac cycles of echocardiographic images and analyzing the results, this study can verify the healthy hearts' contraction and diastole synchronicity of ventricular and atrium, then judge whether the heart's function is healthy.