Evaluation of Limited Angle Reconstruction Methods for Dose Reduction in Cone Beam Computed Tomography Systems

• Main Authors: Su-May, 蘇楣
• Other Authors: Yu-Hua Fang
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/gc867t

碩士 === 國立成功大學 === 生物醫學工程學系 === 105 === Computed tomography (CT) is already an indispensable tool and widely applied in clinical diagnosis, so is cone beam computed tomography (CBCT) in maxillofacial, limb and dental applications. Radiation dose reduction of CBCT remains to be an active topic of research. In the literature, image reconstruction with limited angles can directly reduce the radiation dose, and the iterative reconstruction algorithms have superior performance on limited angles reconstruction. Although there are many kinds of iterative reconstruction algorithms, there is no systematic and objective comparison of these iterative reconstruction algorithms. In this work, we compared five common iterative reconstruction algorithms to the non-iterative algorithm, the L. A. Feldkamp, L. C. Davis, and J. W. Kress (FDK) algorithm, which is common in clinical use, to evaluate which iterative reconstruction method is preferred in limited angle reconstruction. Methods: To set up a prototype CBCT system, we integrated three hardware components: X-ray tube, flat-panel detector and a motor with in-house controlling software. Acquired projection data were used to reconstruct image volumes with tomographic iterative GPU-based reconstruction toolbox (TIGRE), an open source toolbox of various image reconstruction methods. A micro-CT bar pattern phantom and anolis sagrei cadaver were used as irradiated objects. We analyzed the reconstructed images with image quality criteria of signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), universal quality Index (UQI ), root mean of the squared error (RMSE) and full width at half maximum (FWHM). Results: Inspecting the coronal and sagittal planes of the reconstructed results, MLEM, OS-SART and OSC-TV were able to maintain a good image quality with limited angle reconstruction. In the micro-CT bar pattern phantom results, the top three best-performing algorithms are OS-SART, OSC-TV, and MLEM. In the anolis sagrei cadaver results, the best algorithms are OSC-TV, MLEM, and OS-SART. Conclusion: In the iterative reconstruction algorithms tested, MLEM, OS-SART and OSC-TV have superior performance than other algorithms. Since their performances in image quality vary between imaged objects, the choice of reconstruction algorithm may be task-dependent and requires further investigation. This study provides future studies with a framework of testing and evaluation for determining the ideal iterative reconstruction method in the CBCT system development.