Effect of compressed sensing reconstruction on target and organ delineation in cone-beam CT of head-and-neck and breast cancer patients.

Abstract

Compressed sensing (CS) based cone-beam computed tomography (CBCT) reconstruction techniques have been shown to improve image quality. This study was to investigate possible improvements of CBCTCS on manual delineation uncertainties of targets and organs-at-risk. Eight H&N and eight breast cancer patients were selected. Each H&N or breast cancer patient had planning-CT (pCT), repeat-CT (rCT), and CBCT reconstructed by both Feldkamp (CBCTFDK) and compressed sensing methods. On each scan, targets and organs-at-risk were delineated by a radiation oncologist. The impact of reconstruction technique was quantitatively assessed by dice similarity coefficient (DSC) and the shortest perpendicular distance (SPD) between contours of two corresponding scans. The mean CBCTCS-to-rCT DSC was 7.2% and 8.0% bigger than the CBCTFDK-to-rCT for H&N and breast cancer patients respectively. The mean CBCTCS-to-rCT SPD was 16.6% and 25.4% smaller than CBCTFDK-to-rCT SPD. Due to anatomical changes, delineation accuracy reduced in reference to pCT, but no time trend was observed in CBCT based delineation accuracy in reference to rCT. This study demonstrated that CBCTCS has the potential to improve delineation accuracy in H&N and breast cancer patients over CBCTFDK, and CBCTCS thus has potential for adaptive radiotherapy.