A pilot study evaluating 2D antiscatter grid based scatter suppression for quantitative cone beam CT in radiotherapy.

BACKGROUND

Cone beam CT (CBCT) images often suffer from poor CT number accuracy, which adversely affects the use of CBCT in online dose calculations, structure segmentation for plan adaptations, and CBCT-based response assessment during radiotherapy. A major cause of inferior CBCT image quality is X-ray scatter. To address this challenge and improve CBCT fidelity, a dedicated two-dimensional (2D) antiscatter grid was developed and its impact on CBCT image quality was evaluated in a prospective trial. This new approach is referred to as quantitative CBCT (qCBCT).

METHODS

A total of 31 patients undergoing radiotherapy to the prostate, pelvis, abdomen, or head and neck regions were enrolled and scanned using both a standard-of-care clinical CBCT protocol and the qCBCT, which employed a new 2D antiscatter grid prototype in a C-shaped CBCT-linac. Both scans were performed using identical acquisition parameters on the same CBCT system. For the qCBCT scan, the standard 1D antiscatter grid was exchanged with the 2D antiscatter grid. Clinical CBCT scans were reconstructed using standard and iterative reconstruction algorithms. CT number errors, artifact amplitudes, and contrast-to-noise ratio (CNR) in soft tissues were compared between clinical CBCT and qCBCT. Statistical significance of differences between clinical CBCT and qCBCT image quality metrics was evaluated using a linear mixed model.

RESULTS

When compared to clinical CBCT, qCBCT reduced absolute CT number errors and artifact amplitudes by up to 56 % and 53 %, respectively (p < 0.0001). CNR was 20 % higher in clinical CBCT with iterative reconstruction (p < 0.001) and 11 % lower in standard clinical CBCT (p < 0.05), relative to qCBCT. In patients with lateral separation greater than 40 cm, qCBCT yielded up to 62 % lower CT number errors and 60 % lower artifact amplitudes (p < 0.0001). CT number error reduction by qCBCT reached 72 % in the lower neck region (p < 0.0001).

CONCLUSION

This pilot trial showed that hardware-based scatter mitigation with a 2D antiscatter grid significantly improved CT number accuracy in CBCT when compared to clinically available CBCT imaging methods. The improvement in quantitative accuracy was more pronounced for patients with lateral separation greater than 40 cm and in the lower neck region.