Application of a deformable registration technique to investigate breath-hold reproducibility

To develop an objective and less invasive method for evaluating breath-hold status, and to investigate breath-hold reproducibility during voluntary breath-holding irradiation.

We enrolled 20 patients who were treated using a voluntary breath-holding technique. Four or five sessions of cone beam computed tomography (CT) were performed during the radiotherapy session. An image of the lung with respiratory tract was extracted from the CT findings. The rigid registration of subsequent CT findings was completed using the first fraction CT findings. Next, subsequent CT images already subjected to rigid registration were deformed using the first CT images. We compared the differences in the subsequent CT images before and after the deformable registration.

We were able to complete the method to evaluate breath-hold status without having to consider set-up uncertainty, manual processes, visual estimates, or excessive patient cooperation. Tumor dislocation was almost within 3 mm in all directions and deformation was much smaller than 30 % period of free breathing.

This method we have developed may have great potential for the objective verification of breath-hold reproducibility in a less invasive manner. This method assumed that the voluntary breath-hold status was reproducible enough at a clinically satisfactory level.