Image Motion Estimation for Medical Image Applications

• Main Authors: Yu-Chang Hu, 胡育彰
• Other Authors: Kang-Ping Lin
• Format: Others
• Language: en_US
• Published: 2001
• Online Access: http://ndltd.ncl.edu.tw/handle/74313234993160896178

碩士 === 中原大學 === 電機工程研究所 === 89 === In medical image process, cardiac motion estimation is a popular and important research for physicians to diagnose cardiac disease. There are many researches for cardiac motion estimation and left ventricle volume measurement based on X-ray films, MRI images, etc. Therefore, a reasonable and powerful technology will be useful for disease diagnosis. So, the goal of this study is to deal with perfusion images for motion correction of cardiac images and volume measurement of left ventricle. Generally, cardiac perfusion images are based on time parameter extraction and display from a digitized cardiac image series following contrast injection into the left ventricle and right ventricle. Extraction of time parameters from pixel-densograms is a very useful and clinically relevant procedure. This is particularly true when applied to the left ventricle and myocardial, since time parameters indicate the progress of the contract bolus, which in turn reflects the speed and quantify of blood flow. But breath holding is the problem of perfusion images obtained. When patients don’t hold their breath well, it would make the heart moving away from fixed position. It would make us confused by the radioactive medicament concentration for diagnosing. In this study, we will use a tracking process to fix the left ventricle into the same position. Optical flow estimation and block match are two of the most used motion estimation methods. They will be compared how the accuracy they are when we deal with motion correction for perfusion images. Another study is to measure the volume value of left ventricle. A most used technology for volume measurement of left ventricle is segmentation method. Any kinds of segmentation method would have a good results for segmenting to left ventricle and myocardial, but the common problem for left ventricle volume measurement is excluding pop-muscle of left ventricle for measuring. In this study, we use a tracking method to locate the landmarks in all frames of left ventricle volume and use an interpolation process to link landmarks for left ventricle volume measurement. It could solve the problem of excluding pop-muscle and decrease the error measurement of left ventricle volume. As this process applying, it also can decrease the computing time for volume measurement. After all, our purpose of this study is to provide another procedure for motion correction and volume measurement of left ventricle.