(Tuesday, 7/16/2019) 3:45 PM - 4:15 PM
Room: Exhibit Hall | Forum 5
Purpose: To investigate the effect of metal-artifact reduction on an auto-contouring algorithm in computed-tomography images of the head and neck of patients with dental fillings.
Methods: Ten patients with dental fillings were CT scanned (Siemens Somatom Confidence RT Pro) for simulation of treatment planning in radiation therapy of the head and neck. Two sets of images were reconstructed: one without and one with metal-artifact reduction (MAR) for dental fillings. Using a commercial auto-contouring software (Siemens syngo.via VB30), the following anatomical structures were delineated on both data sets: eye globes, lips, mandible, parotids, submandibular glands, and larynx. The results from the two data sets were compared qualitatively against the visible anatomy and quantitatively by the difference in volume of the contours.
Results: As expected from the proximity to dental fillings, the lips showed the largest change in contoured volume between MAR and non-MAR images (mean 30%, min -8%, max 116%), followed by the parotids (mean 6%, min -9%, max 46%), and mandible (mean 5%, min -6%, max 32%). Interestingly, also structures distal to the dental filling exhibited variations in volume, for example, the eye globes (mean 5%, min -14%, max 28%). Qualitatively, auto-contours based on MAR images agreed better with the visible anatomy in most cases.
Conclusion: Auto-contouring algorithms are sensitive to image artifacts, in particular those from dental fillings in CT scans of the head and neck. MAR is effective in reducing some of these artifacts and results in more accurate auto-contouring results. By the time of the conference we will have expanded the patient database and included a comparison of auto-contours against physician contours.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
