| Summary: | 碩士 === 國立中央大學 === 機械工程研究所 === 100 === Ultrasound scan is the most common way to detect the disease of liver diseases. Ultrasound imaging is real-time and non-invasive, but its image quality is poor and lack of spatial information. Therefore, subjective interpretation of ultrasound images may cause misdiagnosis. On the other hand, computed tomography images have high quality. If CT and US images are scanned at similar respiration and posture status, the fusion of both images will be able to improve diagnosis quality and biopsy accuracy.
Prior to ultrasound scan, the patient''s posture and respiration are monitored and controlled to have similar status as CT scan. Then the grabbed ultrasound images of ribs are processed to find the boundary of the ribs. Using the iterative closet point algorithm, the data of boundary points are then applied to complete surface registration with reconstructed 3D model of rib’s CT images and to obtain the transformation matrix between the two image frames.
Experiments of image registration error assessment are performed with an anthropomorphic phantom and a human patient. The edge information of captured characteristics ultrasound images are transformed to CT image frame and compared with the same characteristics on the CT images to calculate their position errors. The results show that position error of anthropomorphic phantom experiment at the xy plane is 2.32mm and the z axis is 4.48mm while that of human patient is approximate 4mm. Since there is clear angular error, press of ultrasound probe on the patient during ultrasound scan may be the factor caused angular error.
|
|---|
