Accuracy of contour propagation from planning computed tomography to iterative cone-beam computed tomography using a deformable image registration algorithm for assisting head and neck radiotherapy

Background

The deteriorated image quality of cone-beam computed tomography (CBCT) reduces the accuracy of contour propagation. We investigated the accuracy of contour propagation from planning CT (pCT) to iterative CBCT (iCBCT) using deformable image registration and compared it with that of replanning CT (reCT) and Feldkamp–Davis–Kress algorithm-based CBCT (FDK-CBCT). No report exists regarding iCBCT improving the accuracy of this technique for images of the head and neck region.

Methods

We included 29 patients who underwent radiotherapy for head and neck cancer. ReCT and CBCT were performed on the same day. The gross tumor volume (GTV) and organs at risk, including the brain stem, spinal cord, mandible, parotid glands, submandibular glands, and larynx, were manually contoured by radiation oncologists on pCT and reCT images. Contour propagation was performed using MIM software. Manually delineated contours on reCT images and deformably generated contours on reCT, FDK-CBCT, and iCBCT images were compared to determine the accuracy of contour propagation using the Dice similarity coefficient (DSC), mean distance to agreement (MDA), and Hausdorff distance (HD).

Results

The mean DSC values for all contoured organs were 0.84 across reCT, FDK-CBCT and iCBCT. A mean DSC value of >0.8 was observed for all organs, except for the larynx and the GTV. The MDA was <1.5 mm for all organs and images, whereas the HD value showed a variation of >3 mm.

Conclusion

The results demonstrated no statistically significant difference in contour propagation from pCT to iCBCT compared to pCT to reCT.