EVALUATION OF ARTIFACTS IN EXPERIMENTAL CINE 4D CT ACQUISITION METHODS

Four-dimensional computed tomography (4D CT) has increased the accuracy of radiation treatment planning for patients in whom the extent of target motion is large. 4D CT has become a standard of care for radiation treatment simulation, allowing decreased motion artifacts and increased spatiotemporal localization of anatomical structures that move. However, motion artifacts may still remain. These artifacts, or artificial anatomic spatial distributions, add a systematic uncertainty to the treatment process and limit the accuracy of lung function images derived from CT. We proposed to reduce the motion artifacts in cine 4D CT by using three novel investigational 4D CT acquisition methods: (1) oversampling the data acquired, (2) gating the x-ray beam with breathing irregularities, and (3) rescanning areas of the clinical standard 4D CT associated with high breathing irregularities. These experimental acquisitions were tested through a protocol approved by the institutional review board with 18 patients with a primary thoracic malignancy receiving a standard 4D CT scan for radiation treatment simulation. The artifact presence in all 4D CT scans was assessed by an automated artifact quantification metric. This artifact metric was validated by a rigorous receiver operating characteristic (ROC) analysis using a high-quality dataset derived from a group of expert observers who reached a consensus decision on the artifact frequency and magnitude for each of 10 clinical 4D CT scans from patients with primary thoracic cancer. The clinical and experimental 4D CT acquisitions from the 18 patients on the protocol were post-processed by the clinical standard of phase sorting and by an experimental phase sorting that incorporated the validated artifact metric. The 4D CT acquisition and processing method judged to be the most improved was the oversampling acquisition with the experimental sorting. The reproducibility of this improved method was tested on a second distinct cohort of 10 patients with a primary thoracic malignancy. Those patients received a clinical phase-sorted 4D CT immediately followed by three independent oversampling acquisitions, processed by the experimental sorting method and evaluated using the artifact metric. The experimental-sorted oversampling acquisition produced a statistically significant artifact reduction (27% and 28% per cohort) from the phase-sorted clinical standard acquisition