Development and application of accurate system calibration in cone beam micro computed tomography system

• Main Authors: Shang-Siou Yang, 楊尚修
• Other Authors: Ching-Huang Lin
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/63672631496209341886

碩士 === 國立雲林科技大學 === 電子與光電工程研究所碩士班 === 101 === Cone-beam computed tomography (CBCT) has been widely used in various applications in recent years. However, if the whole system is not well calibrated, it will influence the image quality seriously. The purpose of this study was to develop an automatic calibration system by a home-made phantom’s projection image analysis that can reduce the errors and obtain more precise projection images. Moreover, this study also develops two the image analysis methods: ratio analysis and independent components analysis (ICA) for dual-energy radiography and CT (DECT) in order to differentiate the different components from the mixed images. In this study, we used a homemade calibration phantom to implement the system calibration at the middle axis (rotational axis) and the ellipse parameters to obtain the system error values. However, it only works under the condition that the detector is parallel to rotational axis without any tilt. The calibration also combined the auto-slider to achieve minimum error. In DECT images, we used ratio analysis and fast independent components analysis (Fast ICA) to analyze the projection image and reconstructed image respectively to differentiate the material compositions. After we obtained the calibration parameters, we did the projection data correction before image reconstruction and the signal to noise ratio (SNR) was improved from 5.241 to 9.7906 after calibration. The linearity was improved from 0.99613 to 0.99959, and the full width at half maximum (FWHM) of point spread function (PSF) was declined from 246.07μm to 239.02μm. The results demonstrated that it could improve the image quality. In addition, in the analysis of dual energy radiography and DECT, the components could be separated by ratio analysis from the projection images and reconstructed images, respectively, but the Fast ICA could not differentiate clearly in both 2D and 3D cases.