Image-guided radiotherapy for lung cancer: Respiration-correlated (cone-beam) CT to verify tumor position and motion characteristics during treatment delivery

Abstract

Purpose/ObjectiveIn radiotherapy of lung cancer it is essential that the average position and motion characteristics of the tumor during treatment delivery are identical to those during treatment preparation. In this work we present a method to verify the tumor position and motion characteristics by comparing respiration-correlated cone-beam CT scans (RC-CBCT) made on the treatment machine with respiration-correlated CT scans (RC-CT) made for treatment preparation.Materials/MethodsRC-CT scans were acquired using a helical scan with 0.8 sec. tube rotation, a pitch of 0.3, and a slice thickness of 3 mm, resulting in a slice distance of 0.9 mm. During scanning, the breathing signal of the patient was recorded using a small thermometer under the nose. Using the thermometer signal, reconstructed CT slices were sorted by phase in the breathing cycle. Slices made at the same phase were combined to reconstruct the thorax at this phase. By interpolation, a total of 32 CT scans were reconstructed covering the whole breathing cycle.RC-CBCT scans were acquired prior to delivery of the first radiation fraction, using an Elekta Synergy XVI system mounted on an Elekta linac. A total of 760 projections was acquired over a full gantry rotation. Since one projection covers the whole lung, the observed position of the diaphragm in each projection could be used as indicator for the breathing phase. During cone-beam reconstruction, only projections at a certain breathing phase were selected, resulting in a full 3-D reconstruction of the thorax at this phase. In this way, 8 CBCT scans (consisting of 95 projections each) were reconstructed, covering the whole breathing cycle.The RC-CBCT scan was fused with the RC-CT scan in an off-line procedure. Fusion was accomplished using a chamfer matching procedure on the stationary bony anatomy (mainly vertebrae) in the average RC-CT (averaged over all phases) and the average RC-CBCT (using all 760 projections). Tumor motion in RC-CT and RC-CBCT was quantified by defining a clip box around the tumor manually in each phase followed by grey-value matching, thus determining the translation of the tumor from a particular phase to the position in a reference phase. To estimate the accuracy of the tumor match, the procedure was repeated several times, each time with a different phase as the reference. Differences in absolute tumor position (with respect to stationary bony anatomy) were assessed by comparing the position of the center of gravity of the tumor in the RC-CT and RC-CBCT scans.ResultsAt this moment, the above procedure has been applied to three patients. The accuracy of the RC-CT - RC-CBCT match was similar to a CT-CT match using a conventional scanning technique, i.e. better than 1 millimeter. The accuracy of the tumor match between different phases was 0.2 mm (1SD) in left-right, 0.7 mm (1SD) in cranial-caudal, and 0.3 mm (1SD) in anterior-posterior direction. All three patients showed a surprisingly good agreement between tumor motion and tumor position in RC-CT and RC-CBCT scans, indicative for a reproducible breathing pattern. The difference in amplitude of the tumor motion along the three principal axes was less than 2 mm. Differences in the position of the center of gravity with respect to the bony anatomy was less than 1.5 mm (vector length).ConclusionsWe developed a method to analyze and compare in detail tumor position and tumor motion in the treatment preparation phase. Using RC-CBCT, these aspects were validated on the treatment machine just prior to irradiation. Fusion of RC-CT and RC-CBCT can be done with high accuracy and tumor motion and position can be determined accurately.RC-CT will be used as basis for treatment planning and a protocol to correct for observed differences in tumor position and tumor motion between treatment delivery and treatment planning is being designed Purpose/ObjectiveIn radiotherapy of lung cancer it is essential that the average position and motion characteristics of the tumor during treatment delivery are identical to those during treatment preparation. In this work we present a method to verify the tumor position and motion characteristics by comparing respiration-correlated cone-beam CT scans (RC-CBCT) made on the treatment machine with respiration-correlated CT scans (RC-CT) made for treatment preparation. In radiotherapy of lung cancer it is essential that the average position and motion characteristics of the tumor during treatment delivery are identical to those during treatment preparation. In this work we present a method to verify the tumor position and motion characteristics by comparing respiration-correlated cone-beam CT scans (RC-CBCT) made on the treatment machine with respiration-correlated CT scans (RC-CT) made for treatment preparation. Materials/MethodsRC-CT scans were acquired using a helical scan with 0.8 sec. tube rotation, a pitch of 0.3, and a slice thickness of 3 mm, resulting in a slice distance of 0.9 mm. During scanning, the breathing signal of the patient was recorded using a small thermometer under the nose. Using the thermometer signal, reconstructed CT slices were sorted by phase in the breathing cycle. Slices made at the same phase were combined to reconstruct the thorax at this phase. By interpolation, a total of 32 CT scans were reconstructed covering the whole breathing cycle.RC-CBCT scans were acquired prior to delivery of the first radiation fraction, using an Elekta Synergy XVI system mounted on an Elekta linac. A total of 760 projections was acquired over a full gantry rotation. Since one projection covers the whole lung, the observed position of the diaphragm in each projection could be used as indicator for the breathing phase. During cone-beam reconstruction, only projections at a certain breathing phase were selected, resulting in a full 3-D reconstruction of the thorax at this phase. In this way, 8 CBCT scans (consisting of 95 projections each) were reconstructed, covering the whole breathing cycle.The RC-CBCT scan was fused with the RC-CT scan in an off-line procedure. Fusion was accomplished using a chamfer matching procedure on the stationary bony anatomy (mainly vertebrae) in the average RC-CT (averaged over all phases) and the average RC-CBCT (using all 760 projections). Tumor motion in RC-CT and RC-CBCT was quantified by defining a clip box around the tumor manually in each phase followed by grey-value matching, thus determining the translation of the tumor from a particular phase to the position in a reference phase. To estimate the accuracy of the tumor match, the procedure was repeated several times, each time with a different phase as the reference. Differences in absolute tumor position (with respect to stationary bony anatomy) were assessed by comparing the position of the center of gravity of the tumor in the RC-CT and RC-CBCT scans. RC-CT scans were acquired using a helical scan with 0.8 sec. tube rotation, a pitch of 0.3, and a slice thickness of 3 mm, resulting in a slice distance of 0.9 mm. During scanning, the breathing signal of the patient was recorded using a small thermometer under the nose. Using the thermometer signal, reconstructed CT slices were sorted by phase in the breathing cycle. Slices made at the same phase were combined to reconstruct the thorax at this phase. By interpolation, a total of 32 CT scans were reconstructed covering the whole breathing cycle. RC-CBCT scans were acquired prior to delivery of the first radiation fraction, using an Elekta Synergy XVI system mounted on an Elekta linac. A total of 760 projections was acquired over a full gantry rotation. Since one projection covers the whole lung, the observed position of the diaphragm in each projection could be used as indicator for the breathing phase. During cone-beam reconstruction, only projections at a certain breathing phase were selected, resulting in a full 3-D reconstruction of the thorax at this phase. In this way, 8 CBCT scans (consisting of 95 projections each) were reconstructed, covering the whole breathing cycle. The RC-CBCT scan was fused with the RC-CT scan in an off-line procedure. Fusion was accomplished using a chamfer matching procedure on the stationary bony anatomy (mainly vertebrae) in the average RC-CT (averaged over all phases) and the average RC-CBCT (using all 760 projections). Tumor motion in RC-CT and RC-CBCT was quantified by defining a clip box around the tumor manually in each phase followed by grey-value matching, thus determining the translation of the tumor from a particular phase to the position in a reference phase. To estimate the accuracy of the tumor match, the procedure was repeated several times, each time with a different phase as the reference. Differences in absolute tumor position (with respect to stationary bony anatomy) were assessed by comparing the position of the center of gravity of the tumor in the RC-CT and RC-CBCT scans. ResultsAt this moment, the above procedure has been applied to three patients. The accuracy of the RC-CT - RC-CBCT match was similar to a CT-CT match using a conventional scanning technique, i.e. better than 1 millimeter. The accuracy of the tumor match between different phases was 0.2 mm (1SD) in left-right, 0.7 mm (1SD) in cranial-caudal, and 0.3 mm (1SD) in anterior-posterior direction. All three patients showed a surprisingly good agreement between tumor motion and tumor position in RC-CT and RC-CBCT scans, indicative for a reproducible breathing pattern. The difference in amplitude of the tumor motion along the three principal axes was less than 2 mm. Differences in the position of the center of gravity with respect to the bony anatomy was less than 1.5 mm (vector length). At this moment, the above procedure has been applied to three patients. The accuracy of the RC-CT - RC-CBCT match was similar to a CT-CT match using a conventional scanning technique, i.e. better than 1 millimeter. The accuracy of the tumor match between different phases was 0.2 mm (1SD) in left-right, 0.7 mm (1SD) in cranial-caudal, and 0.3 mm (1SD) in anterior-posterior direction. All three patients showed a surprisingly good agreement between tumor motion and tumor position in RC-CT and RC-CBCT scans, indicative for a reproducible breathing pattern. The difference in amplitude of the tumor motion along the three principal axes was less than 2 mm. Differences in the position of the center of gravity with respect to the bony anatomy was less than 1.5 mm (vector length). ConclusionsWe developed a method to analyze and compare in detail tumor position and tumor motion in the treatment preparation phase. Using RC-CBCT, these aspects were validated on the treatment machine just prior to irradiation. Fusion of RC-CT and RC-CBCT can be done with high accuracy and tumor motion and position can be determined accurately.RC-CT will be used as basis for treatment planning and a protocol to correct for observed differences in tumor position and tumor motion between treatment delivery and treatment planning is being designed We developed a method to analyze and compare in detail tumor position and tumor motion in the treatment preparation phase. Using RC-CBCT, these aspects were validated on the treatment machine just prior to irradiation. Fusion of RC-CT and RC-CBCT can be done with high accuracy and tumor motion and position can be determined accurately.