Detection of anatomical changes using two‐dimensional x‐ray images for head and neck adaptive radiotherapy

摘要

Purpose To develop a system for detecting anatomical changes using two‐dimensional (2D) x‐ray images. Methods Ten patients with head and neck cancer were retrospectively analyzed using 2D x‐ray and cone‐beam computed tomography (CBCT) images. The 2D x‐ray images were acquired daily, whereas the CBCT images were acquired weekly during the treatment period. The developed system imported the 2D x‐ray images obtained on the initial treatment day and on another day, and thereafter converted them into the water equivalent thickness (WET) using the conversion table. The difference between the WET images for the first and other treatment days (ΔWET) was calculated as the quantitative value for anatomical changes and visualized to recognize the anatomical change location. We compared ΔWET and the difference in the lateral neck distance (ΔLND) on the corresponding CBCT images. ΔLND was used as the ground truth for anatomical changes. ΔWET and ΔLND were measured at the first cervical vertebra (C1) and the tumor center (TC). C1 and TC were selected to observe the volume changes in the parotid gland and tumor, respectively. Sensitivity and specificity were calculated to evaluate the performance of the 2D‐WET system. The cut‐off values of WET and LND were set to 2–10?mm. Furthermore, intensity‐modulated proton therapy (IMPT) plans for six patients with rescan CT images were generated. The IMPT plans on the rescan CT images were compared to the original plans on simulation CT using the dosimetric parameters for the target and the organs at risk. Results The mean differences between ΔWET and ΔLND for C1 and TC were ?0.62 ± 1.66?mm and ?0.93 ± 1.28?mm (mean ± 1 SD), respectively. ΔWET in the proposed system was in good agreement with ΔLND using the CBCT images. In the sensitivity and specificity results for C1 and TC with cut‐off values from 2 to 10?mm, the sensitivity was?>85 for all cut‐off values, while the specificity was?>90 at 5–10?mm and?<90 at less than 5?mm. The average ΔWET at the time of replanning was 12.8?mm which resulted in maximum dose increase in the spinal cord D1cc by 8.4?Gy, the parotid gland D50 by 26.6?Gy, and the oral cavity D50 by 23.2?Gy. Conclusions We developed a new system for detecting anatomical changes using 2D x‐ray images. The developed system with ΔWET showed an agreement with ΔLND at C1 and TC with an average difference of less than 1?mm. ΔWET detected anatomical changes with high sensitivity and specificity with a cut‐off value of 5–10?mm. This system can monitor daily anatomical changes without causing high exposure to patients and requiring any inefficient work, and it can be applied to daily online adaptive proton beam therapy and triggered adaptive radiotherapy.