Speaker
Description
Accurate scan geometry is essential for high-quality three-dimensional reconstruction in cone-beam computed tomography (CBCT) [1]. However, residual geometric misalignments frequently persist even after careful mechanical alignment, resulting in image degradation such as blurring and structural distortion. Conventional calibration approaches typically rely on dedicated phantoms, limiting their practicality in routine or industrial environments. In this study, a phantom-free geometry calibration framework is proposed to estimate detector misalignment parameters directly from projection data. As illustrated in Fig. 1, the method focuses on estimating transverse and longitudinal detector shifts (δu and δv) relative to the ideal geometry. The calibration problem is formulated as an optimization task, and Bayesian optimization is employed to efficiently explore the parameter space by modeling the objective function as a gaussian process [2]. This approach enables robust estimation of near-optimal parameters with a limited number of evaluations, thereby reducing computational burden while maintaining accuracy. The proposed method was validated through both numerical simulation and experimental studies (Fig. 2). A 3D foam-structure phantom was used to provide a controlled simulation environment, while a tissue-equivalent phantom was employed for realistic experimental validation. In simulation, geometric misalignment resulted in significant structural distortion, whereas the optimized parameters effectively restored the reference geometry (Fig. 3). Quantitative analysis demonstrated a substantial reduction in root mean square error (RMSE) of up to 94.9%, along with improved agreement in line profiles. In experimental results, the estimated geometric parameters successfully compensated for misalignment effects, leading to enhanced structural fidelity and reduced blurring in reconstructed images (Fig. 4). Enlarged views further confirmed improved edge definition and suppression of distortion artifacts. These results demonstrate that the proposed phantom-free calibration framework provides an effective and practical solution for correcting geometric misalignment in CBCT without requiring dedicated calibration phantoms, with strong potential for application in industrial and clinical imaging systems.