Usability of Deformable Image Registration for Adaptive Radiotherapy in Head and Neck Cancer and an Automatic Prediction of Replanning

Abstract

Deformable image registration (DIR) has been an important component in adaptive radiotherapy (ART). Our goal was to examine the accuracy of ART using the dice similarity coefficient (DSC) and to determine the optimal timing of replanning. A total of 22 patients who underwent volume modulated arc therapy (VMAT) for head and neck (H&N) cancers were prospectively analyzed. The planning target volume (PTV) was to receive a total of 70 Gy in 33 fractions. A second planning CT scan (rescan) was performed at the 15th fraction. The DSC was calculated for each structure on both CT scans. The continuous variables to predict the need for replanning were assessed. The optimal cut-off value was determined using receiver operating characteristic (ROC) curve analysis. In the correlation between body weight loss and DSC of each structure, weight loss correlated negatively with DSC of the whole face (r s = −0.45) and the face surface (r s = −0.51). Patients who required replanning tended to have experienced rapid weight loss. The threshold DSC was 0.98 and 0.60 in the whole face and the face surface, respectively. Patients who showed low DSC in the whole face and the face surface required replanning at a significantly high rate (P < 0.05 and P < 0.01). Weight loss correlated with DSC in both the whole face and the face surface (P < 0.05 and P < 0.05). The DSC values in the face predicted the need for replanning. In addition, weight loss tended to correlate with DSC. DIR during ART was found to be a useful tool for replanning.

Early Prediction of Planning Adaptation Requirement Indication Due to Volumetric Alterations in Head and Neck Cancer Radiotherapy: A Machine Learning Approach

“…However, ART’s complicated computations (where clinicians have to constantly analyze new scans and create new plans), significantly increase the overall workload and resource utilization burden, underlying the need for an efficient and cost-effective approach for the identification of the patients that will benefit from ART [ 11 ]. To that end, a large number of studies have investigated the statistical relationship between the need for ART and image-based characteristics [ 12 , 13 ]. Nevertheless, the development of early prediction multivariate models for ART application presents several challenges as a consequence of volume-related discrepancies, resulting from anatomical alterations in cancer patients treated with RT [ 14 , 15 ].…”

Variations of the Dose Distribution Between CT- and CBCT-based Plans for Oropharyngeal Cancer

“…Using CBCT obtained from the on-board imager (OBI), the patient's contour can be delineated and its shape and extent changes can be determined using deformable image registration (DIR) from original planning (9)(10)(11). For head and neck cancer patients, weight loss and setup accuracy are important, as these influence the dose distribution during IMRT (12)(13)(14). Furthermore, setup errors in the neck region of patients are generally greater compared to those occurring in the head region during the course of radiotherapy (15).…”