Dual-scan conformal cone-beam CT for targeted image-quality improvement using dynamic collimation
Yuxiang Liu1, Siqi Yuan1, Xin Feng1
1National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China.
Background:
Cone-beam computed tomography (CBCT) is routinely used for image guidance in radiation therapy, but conventional full-field CBCT acquisition may expose anatomically irrelevant regions and generate substantial scatter, which degrades soft-tissue contrast and limits the accuracy of target localization and adaptive radiotherapy workflows. A clinically practical imaging strategy should improve image quality in the region of clinical interest while preserving sufficient full-field anatomical information for patient setup and dose-related assessment.
Purpose:
To improve image quality within clinically relevant regions without increasing the total imaging dose, this study developed a dual-scan conformal cone-beam CT method using dynamic collimation.
Methods:
A dual-scan conformal acquisition strategy using dynamic collimation was designed. Dynamic collimation was defined as angle-dependent shaping of the X-ray field around a preplanned target region; it was implemented with a multi-leaf collimator model in the digital phantom and patient-data simulations and with the movable kV collimator in the physical phantom measurement. Collimator positions were calculated from forward projections of the target region, and a target-region-based intensity-optimization model was used to allocate more photons to projection angles and detector regions contributing to the target region. The dual-scan protocol, consisting of a low-dose full-field-of-view (FOV) scan and a higher-dose conformal scan, was reconstructed using a regularized weighted least-squares algorithm with anisotropic total-variation regularization. The method was evaluated using a FORBILD head phantom, patient planning CT datasets, and an anthropomorphic head phantom measured on a Varian Edge on-board imaging system.
Results:
The proposed method achieved the best overall image quality among the compared methods. In the FORBILD phantom, for target region 1 (VOI1), SSIM increased from 0.702 to 0.855, CNR increased from 0.514 to 1.463, and SNR1/SNR2 increased from 33.8/32.8 to 113.1/114.6 compared with conventional FDK reconstruction. In the prostate cancer case, CNR increased from 0.44 to 1.94, and SNR1/SNR2 increased from 25.6/23.3 to 74.6/82.4. In the measured anthropomorphic head phantom, CNR improved from 1.036 to 2.379, and SNR1/SNR2 improved from 42.8/50.1 to 89.8/98.0.
Conclusion:
The proposed dual-scan conformal CBCT strategy using dynamic collimation improves local image quality in the clinically relevant target region while maintaining full-FOV information through the two-scan protocol. This method may provide a practical route toward patient-specific CBCT guidance for adaptive radiotherapy.
